Med 21255

Fragment Informatics andComputational Fragment-Based Drug

Design: An Overview and Update

Chunquan Sheng and Wannian Zhang

Department of Medicinal Chemistry, School of Pharmacy, Second Military Medical University, 325 GuoheRoad, Shanghai, 200433, People’s Republic of China

Published online 19 March 2012 in Wiley Online Library (wileyonlinelibrary.com).DOI 10.1002/med.21255

�

Abstract: Fragment-based drug design (FBDD) is a promising approach for the discovery and optimiza-tion of lead compounds. Despite its successes, FBDD also faces some internal limitations and challenges.FBDD requires a high quality of target protein and good solubility of fragments. Biophysical techniquesfor fragment screening necessitate expensive detection equipment and the strategies for evolving fragmenthits to leads remain to be improved. Regardless, FBDD is necessary for investigating larger chemical spaceand can be applied to challenging biological targets. In this scenario, cheminformatics and computationalchemistry can be used as alternative approaches that can significantly improve the efficiency and successrate of lead discovery and optimization. Cheminformatics and computational tools assist FBDD in a veryflexible manner. Computational FBDD can be used independently or in parallel with experimental FBDDfor efficiently generating and optimizing leads. Computational FBDD can also be integrated into eachstep of experimental FBDD and help to play a synergistic role by maximizing its performance. This reviewwill provide critical analysis of the complementarity between computational and experimental FBDD andhighlight recent advances in new algorithms and successful examples of their applications. In particular,fragment-based cheminformatics tools, high-throughput fragment docking, and fragment-based de novodrug design will provide the focus of this review. We will also discuss the advantages and limitations ofdifferent methods and the trends in new developments that should inspire future research. C© 2012

Wiley Periodicals, Inc. Med. Res. Rev., 33, No. 3, 554–598, 2013

Key words: computational fragment-based drug design; fragment informatics; fragment docking;fragment-based de novo design

Correspondence to: Wannian Zhang or Chunquan Sheng, Department of Medicinal Chemistry, School of Phar-macy, Second Military Medical University, 325 Guohe Road, Shanghai 200433, People’s Republic of China.E-mail: [email protected] or [email protected]

Medicinal Research Reviews, 33, No. 3, 554–598, 2013C© 2012 Wiley Periodicals, Inc.

FRAGMENT INFORMATICS AND COMPUTATIONAL FBDD � 555

1. INTRODUCTION

The identification of small molecules that selectively bind to a biological target is the key step indrug discovery. High-throughput screening (HTS) is a routine method for hit or lead discoveryin the pharmaceutical industry.1 It has proven to be effective in many research programs,particularly with improved lead-like libraries. Although HTS has produced a number of leadmolecules for various drug targets, it also has several limitations. A HTS campaign usuallyscreens 106–107 compounds, which only cover a small portion of drug-like chemical space(about 106 molecules2). Moreover, the hit rate of HTS is generally low and the resultingleads are difficult to be optimized into drug-like candidates, because many of them have largemolecular weights (MWs) and are strongly hydrophobic.3 In this context, fragment-based drugdesign (FBDD) is becoming an alternative approach for drug discovery.4

Taking advantages of both random screening and structure-based drug design (SBDD),FBDD constructs novel lead structures from small molecular fragments. Since the introductionof the “SAR by NMR” method in 1996,5 FBDD has become a practical and promisingtool in drug discovery.6 The workflow of a FBDD study is depicted in Figure 1. The first

Figure 1. The complementarity between computational and experimental FBDD.

Medicinal Research Reviews DOI 10.1002/med

556 � SHENG AND ZHANG

part of FBDD is to identify weak to moderate binders of the desired target by fragmentscreening. The libraries for fragment screening contain hundreds to thousands of small andlow MW fragments, which are screened at high concentration.7 Because the binding affinitiesof fragments are relatively weak (5 mM–1 μM), highly sensitive detection methods have beendeveloped for this purpose.8 The biophysical techniques for fragment screening mainly includenuclear magnetic resonance (NMR),9 mass spectroscopy (MS),10, 11 X-ray crystallography,12

and surface plasmon resonance (SPR) spectroscopy.13, 14 Then, various optimization strategiescan be used to increase the affinity and drug likeness of fragment hits to evolve them into high-quality leads. The hit-to-lead optimization process may involve a combination of fragmentlinking, fragment evolution, fragment optimization, and fragment self-assembly, which is oftenguided by the structural information of the target-fragment complex.15 The lead optimizationstage is technically similar to that of conventional SBDD. More than ten clinical candidates havebeen generated by the FBDD strategy.4 In 2011, the B-Raf inhibitor vemurafenib (Zelboraf),16

the first FBDD derived drug, was approved by the FDA for the treatment of melanoma. It tookonly 6 years from concept to approval for vemurafenib. Inspired by these encouraging results,FBDD is attracting more and more attention from both the pharmaceutical industry and theacademic community.17–19

Compared to HTS, FBDD has several advantages, including generation of higher chemicaldiversity (sampling a larger chemical space), higher hit rates, and higher ligand efficiency(LE = –log IC50 / number of heavy atoms).20 However, current FBDD approaches also facesome internal limitations and challenges. First, FBDD methods still cover a small fraction ofthe total diversity space. It is estimated that a library of 103 fragments can typically sample thechemical diversity space of 109 molecules. Although the combinatorial advantage of FBDDprovides a significant increase in diversity space relative to HTS, exploring a larger region ofdrug-like space is still needed. Second, current fragment screening methods demand significantamounts, purity, solubility, and suitability of target proteins for labeling or crystallization.Although progress has been made, the successful application of FBDD to membrane proteins(e.g. G-protein coupled receptors, GPCR) remains a significant challenge.21 Moreover, FBDDis more suitable for certain classes of targets whose binding site often consists of multipledistinctive subsites (such as kinases) as individual fragments may occupy different subsites andthey can be joined later into complete molecules. On the other hand, the process of fragmentoptimization is often guided by structure-based design, which is difficult to be applied totargets whose structures are unknown. Third, most of the FBDD methods do not take ligandspecificity or selectivity into account. Although there are good examples of selective fragmentstargeted to kinases, the methodologies of FBDD need to be improved to efficiently identifyfragments that bind to the sites responsible for target specificity.22 Fourth, the geometries andkey interactions of the original fragment hits may be changed when they are evolved intolead compounds.23 New methods should be developed to efficiently select proper linkers tobridge fragments, find proper groups (fragments) to be added to the initial fragment hits, andpredict the binding mode of the newly generated molecules. Last, the techniques of FBDDoften require specialized equipment and specific expertise,24 which limits the broad applicationof FBDD.

Cheminformatics and computational approaches provide an alternative to the experimentalFBDD methods. The incorporation of computational methods into the FBDD process can sig-nificantly improve the efficiency and success rate of lead discovery and optimization. Moreover,the low-throughput nature of experimental FBDD makes computational tools an attractive wayto explore larger commercially available fragment databases. Computational chemistry toolscan significantly improve the efficiency of each step of FBDD, such as fragment library design,active site characterization, fragment hit discovery, and hit-to-lead-to-candidate optimization.Several reviews have covered the topic of computational FBDD approaches.25–32 Our goal here



is to highlight recent advances of new algorithms, analyze their advantages and limitations, anddiscuss the trends to inspire future research. We will first provide a comprehensive overviewof the complementarity between computational and experimental FBDD. Then, a broad setof cheminformatics and computational FBDD tools as well as their successful applicationswill be discussed in detail. In particular, we will focus on recent advances of fragment-basedcheminformatics tools, high-throughput fragment docking, and fragment-based de novo drugdesign.

2. OVERVIEW OF THE COMPLEMENTARITY BETWEEN COMPUTATIONALAND EXPERIMENTAL FBDD

Computational chemistry provides complementary methods for experimental FBDD, and hasassisted the implementation of FBDD in an efficient and cost-effective manner. Computationalapproaches play an important role throughout the process of FBDD (Fig. 1). The constructionof high-quality fragment libraries is the first step in the FBDD process. The library for fragmentscreening should have good diversity to represent drug-like chemical space and also meetcertain criteria of physicochemical properties, solubility and synthetic accessibility.7, 8, 33 Suchproperties can be quickly obtained by computational methods and then used as filters forcommercially available fragment databases. Computational methods are also helpful to removefragments with unwanted chemical groups and incorporate the most frequently occurringfragments from known drugs. The fragments also need to be highly soluble, because theyare screened at a high concentration. Approaches for the prediction of aqueous solubilitymainly include quantitative structure–activity relationships (QSAR) and quantitative structure–property relationships (QSPR) modeling.34, 35

During the fragment screening stage, molecular docking has been used as a prescreen toolto reduce experimental efforts. Virtual fragment screening can also directly yield potent hitswithout using a direct detection technique of experimental FBDD. For the hits identified fromfragment screening, computational approaches (e.g. substructure search and similarity-basedsearch) can be used to facilitate hit expansion and obtain SAR information for a secondaryscreen. For example, a research group from Vertex used the NMR SHAPES36 method to identifyseveral micromolar hits by performing a secondary screen on 500 compounds that were obtainedfrom substructure and similarity searching around the fragment hits.37 Although hit expansionof existing compounds may be a practical approach, it is more desirable to design and synthesizetotally new compounds using fragment hits as “seeds.” Computational analysis of a protein-hitcomplex can provide useful information to prioritize the most promising fragment hits for thesubsequent fragment-to-lead process. Structure-based in silico methods can iteratively assistthe buildup of the fragment hits into a new lead compound that possesses improved potencyand drug likeness. For example, the selection of an appropriate linker to join fragment hits isof great importance to generate high-affinity ligands. The flexibility of the linker is importantfor the binding geometries of the original fragment hits and the binding affinity of the resultingmolecules.38 In this context, computational methods (e.g. de novo drug design algorithms39)are helpful to virtually screen linker libraries. Moreover, de novo drug design methods cannot only significantly aid the assembly of fragment hits into novel compounds, but can alsoautomatically design novel ligands. Molecular docking and molecular dynamics simulationscan efficiently predict the binding affinity and binding pose of the designed ligands, and thusthey are powerful tools for the evolution of fragments into potent leads.40, 41 According toVangreveling’s review,30 31 out of 36 successful FBDD examples used structure-based designtools for fragment-to-lead optimization. Such structure-based approaches are also popular forthe optimization of the lead into a clinical candidate.29



3. FRAGMENT INFORMATICS

What is a fragment? It is difficult to give a precise definition. Generally, fragments are small,low MW and highly soluble molecules that have weak binding affinity with the target protein.Fragments are often used to “build” a larger lead compound with improved biological activity.Also, the term fragment can be regarded as a substructure or structural part of a more complexmolecule. In 2003, Congreve et al. found that fragment hits possessed physicochemical proper-ties that meet the criteria of the “rule of three” (RO3), namely (i) MW ≤ 300 Da; (ii) hydrogenbond donors and acceptors ≤ 3; (iii) LogP ≤ 3.42 Additional physicochemical properties for afragment include three or less rotatable bonds and a polar surface area less than 60 A2. Typically,the MW of fragment hits is in the range of 120–250 and the binding affinity is 30 μM–1 mM.15

More recently, modifications or extensions of these rules have been suggested.43–45 Molecularfragments have long been used as descriptors for chemical similarity searches or diversity anal-ysis and played an important role for chemoinformatics analysis. In addition, fragments havealso been associated with specific biological activities, privileged structural motifs of specifictarget families, and absorption, distribution, metabolism, excretion, and toxicity (ADME/T)profiles.46–48 In the following sections, recent progress of fragment informatics and its impacton FBDD will be described (Fig. 2).

A. Fragmentation Approaches and Fragment Space

Breaking molecules into fragments is the first step in fragment mining or fragment informat-ics analysis. Fragmentation of molecules allows the comparison of molecules using standard

Figure 2. The influence of fragment informatics to FBDD.



cheminformatics approaches. Nowadays, there are various publicly or commercially avail-able databases, such as PubChem,49 eMolecules,50 WOMBAT,51 ZINC,52 WDI,53 Medchem,54

MDDR,55 and CMC,56 that consist of structure, property, and/or biological activity data formillions of small molecules. These easily accessible databases provide useful sources for chem-informatics analysis. Fragments are often obtained by in silico fragmentation of molecularstructures. The average size of fragments and the composition of the fragment population aredetermined by two important parameters: the maximum number of permitted bond deletionsper iteration and the total number of iterations.

A number of well-defined computational fragmentation schemes have been reported. Theycan be mainly classified into substructure methods and building block methods. The substruc-ture approaches treat the fragment as a substructure of the molecules and aim to complete theanalysis of all possible fragments.57 Such methods are not specific to fragments and are oftenused in QSAR or similarity searches. The building block methods mainly focus on chemicallymeaningful fragments and have wide applications in computational FBDD. Predefined break-ing rules are used to dissect molecules into building blocks. For example, building blocks canbe defined as rings, functional groups, side chains, or linkers. Our group has decomposed theMDDR database55 into rings, linkers, and side chains, which can be used to build drug-likefragment libraries.58 Another efficient way of fragmentation is virtual retrosynthesis. RECAPis the most widely used method and employs some common chemical reactions as the rulesto break structures.59 During the process of RECAP fragmentation, the bonds formed by oneof these reactions are cleaved. RECAP has been successfully applied to explore the fragmentspace, analyze drug-like fragments in marketed drugs,60, 61 and construct synthetically feasiblefragment libraries for de novo ligand design.62 More recently, Schulz et al. evaluated six differentcheminformatics tools for the construction of a fragment library.63 An iterative removal proto-col was proven to be the best method to design a diverse fragment library that can maximallyrepresent the commercially available chemical space.

Chemical fragment space means combinations of molecular fragments and their connectionrules. A rather small number of fragments can span a huge space of virtual compounds dueto the “combinatorial explosion.”64 Because the chemical space of fragments is significantlysmaller than that of drug-like molecules,65 good sampling in the fragment space may be achievedby FBDD. An important goal of fragment informatics analysis is to construct drug-like andchemically tractable fragment space that can generate potent active compounds against a largevariety of targets.66

Mauser et al. generated thousand-size fragment space through fragmentation of the WDI200453 and the Medchem0354 databases using the RECAP principle.67 The fragment spacecontained two subsets: a subset containing the most frequently occurring fragments (2039 frag-ments) and a substructure-based diverse subset (1923 fragments). Validation studies revealedthat the two subsets were complementary to each other and that their combination covereda larger part of drug-like chemical space. Tanaka et al. performed network analysis of frag-ment libraries by extracting relatively small compounds from the ZINC database.52 Moreover,an efficient compound-prioritization method was proposed for fragment linking. The varietyof linkers was also shown to be relatively important for molecular diversity when fragmentlinking was performed. More recently, the fragment subset of a large compound database wasanalyzed. Deursen et al. visualized 4.5 million fragments in PubChem and provided importantinformation on the distribution of structural diversity.68

Although RECAP has been widely used, it only covers a very limited number of gen-erally applicable reactions. Other methods constructed fragment spaces that avoid splittingknown molecules by retrosynthetic rules. For example, Cramer’s group from Tripos devel-oped the ChemSpace technology69 and AllChem70 to navigate through known chemistry by theTopomer search methodologies.71 Another study used the Feature Trees Fragment Space Search



(Ftrees-FS) method64, 72 to generate a huge fragment space encoding about 5 × 1011 compoundsbased on established in-house synthetic protocols for combinatorial libraries.73

B. Fragment Frequency Analysis and Fragment Mining

In terms of fragments, a great deal of information can be obtained from existing or virtualcompound databases, such as the distribution of fragments types, their frequency of occurrenceand co-occurrence. Fragment frequency analysis is typically useful for understanding the natureof fragment–activity and fragment–drug-likeness relationships. Cheminformatics analysis ofthe differences of fragments between drugs and nondrugs, or between various classes of drugswill provide medicinal chemists with useful information for prioritizing screening libraries ordesigning drug-like compounds.

There are two types of fragment frequency analysis: occurrence analysis and co-occurrenceanalysis. Occurrence analysis means the characterization of fragment distributions in largedatabases, while co-occurrence analysis is used to compare fragment sets in a pairwise man-ner. Pioneering work on the analysis of drug-like fragments was performed by Bemis andMurcko.60, 61 They identified frequently occurred molecular frameworks and side chains indrug sets selected from the CMC database. A similar strategy has also been applied to variousdatabases to find drug-like fragments.74–77 More recently, Wang and Hou provided an updateon drug-like fragments analysis and identified high-quality fragments for drug design.78 Fre-quencies for three kinds of building blocks (ring system, drug scaffold, and small fragment)were calculated for a FDA-approved drug database (ADDS) and an extended drug dataset(EDDS). Most top fragments were found to be essentially common for both drug datasets.Moreover, there is significant difference in the distribution of chemical fragments between oraldrugs and injectable drugs.79 Therefore, compounds designed by marketed oral drug fragmentsare more likely to have good bioavailability. Fragments that are recurrent in compounds withdifferent activities are relevant as a useful source for the design of multitarget ligands.80, 81

Sheridan et al. used common substructures from the MDDR database to identify fragment re-placements in drug-like molecules82 and fragments that are associated with multiple biologicalactivities.81 Drug-like bioisosteric groups have been identified by the cheminformatics analysisof the frequency of occurrence of organic substituents in more than 3 million molecules.83

Haubertin and Bruneau systematically analyzed one-to-one chemical replacements occurringin an in-house drug-like dataset of AstraZeneca and built a web-based database of historicallyobserved chemical replacements.84 Furthermore, fragment frequency analysis has also beenused to build predictive models for ADME/T prediction.85, 86

Bajorath’s group introduced a new method named MolBlaster that used randomly gen-erated fragment populations to evaluate molecular similarity relationships.87 MolBlaster gen-erates “fragment profiles” of molecules by random deletion of bonds in connectivity tablesand quantitative comparisons using entropy-based metrics. The term “fragment profile” differsfrom “molecular fingerprint,” because it is randomly generated and does not depend on pre-defined structural or property descriptors. Fragment profile can encode sufficient informationfor similarity evaluation, which has been developed into a new tool for ligand-based virtualscreening.88 Furthermore, the same group developed a new methodology to mine and organizerandomly generated molecular fragment populations.89 Unique fragment signatures were iden-tified for molecular sets with similar activities, and then fragment pathways of biologically activemolecules were mapped. The results indicated that compound class-specific information andactivity-specific fragment hierarchies could be obtained from random fragment profiles. Morerecently, Lounkine et al. developed a new approach termed FragFCA to identify molecularfragments and fragment combinations that are specific for compounds having different activity



profiles or that are unique to highly potent molecules.90 FragFCA uses chemically intuitivequeries of varying complexity to systematically identify sets of signature fragments in a flexibleand interactive manner, and is applicable to fragments derived by any fragmentation scheme.

Lameijer et al. reported fragment mining results for the NCI database.91 The database wassplit into more than 60,000 fragments of varying types: ring systems, linkers (2–6 linkers) andsubstituents. After analyzing the fragment occurrence and co-occurrence, the authors identified“chemical cliches” that indicated the most-occurring fragments and frequently co-occurringpairs of fragments. The resulting fragment libraries and correlations can give medicinal chemistsmore ideas about lead optimization, clusters of biologically active structures, and relativelyunexplored parts of chemical space.

Analysis of fragment frequencies in biologically active compounds can also be used to buildinterpretable models for the prediction of activity and target space. A research group from Lillydescribed a simple approach for fragmentation of a literature-based dataset and constructednaive Bayes models for predicting potency in individual kinases by comparing the similarity offragment fingerprints.76 The statistical models had good predictive ability for kinase potency inboth retrospective and prospective tests. Moreover, the comparison of fragment distributions inactive molecules is also useful to assess target similarity. This method based on fragment-derivedsimilarities complements sequence-based comparisons and whole-molecule approaches.

C. Fragment Tree

During the process of lead optimization in drug discovery, medicinal chemists often synthesizea series of analogues with a common or similar core fragment (scaffold). Thus, the structureof a biologically active compound can be dissected into scaffold and substituents (R-groups).In order to find a highly potent compound against a given target, most medicinal chemistryefforts are focused on the variation of the composition of the scaffolds and the substituents.For a small dataset, the importance of the scaffolds and substituents are straightforward andthe SARs can be easily understood. But for large databases, especially the one containing aseries of similar scaffold substructures, it is difficult to clearly elucidate SAR. In this case,organizing the structures in the form of a hierarchical tree is often beneficial to rationalizeSAR. In a hierarchical tree, structures with a common core fragment are arranged in branchesand each node is a substructure (fragment) shared by all of its descendent (smaller) nodes.A well-constructed hierarchical tree can bring insights into which core fragments and whichperipheral substituents are responsible for the activity, toxicity, or other relevant properties.

The methods for building a hierarchical tree are largely dependent on the nature of thedataset. If the compounds were synthesized by linking various substituents with similar corefragments in a stepwise manner, it is easy to construct a fragmentation tree on the basis ofthe synthetic procedures. A descendent hierarchy can also be constructed by using the knowncommon scaffolds as the root fragments. However, scaffolds are not always known ahead of timefor a collection of structurally similar molecules without specific information about commonsubstructures. Thus, algorithms should be developed to determine which parts of a structureare most scaffold-like.

Several methods have been developed for the extraction, identification, and classifica-tion of chemical scaffolds.92–94 Using these approaches, the scaffold tree can be built and thescaffold universe of the dataset can be visualized. More recently, Clark et al. reported new algo-rithms for common scaffold alignment, multiple scaffold detection, and scaffold substructureassignment.95 These methods can address the issues of multiple scaffolds, noncommon scaf-folds, and symmetrical common scaffolds and produce informative data for structure–activityanalysis. Furthermore, the same group described a more informative method for producing



two-dimensional (2D) depiction layout coordinates for each node in a scaffold tree.96 The al-gorithm includes generating a fragment tree, mapping sibling fragments onto each other in anoptimal way, and using this mapping to guide a 2D depiction process. The advantage of theapproach lies in that common ancestor fragments can be depicted and oriented in a consistentway and thus common structural features can be readily evident to medicinal chemists. An in-teractive tool called the scaffold explorer97 differs from other automated scaffold classificationalgorithms in that the scaffolds can be of arbitrary complexity and the user can construct ascaffold tree interactively. Scaffold explorer allows medicinal chemists to accommodate theirintuition and shows good interactivity for mapping SAR across different chemotypes.

The above-mentioned methods for building the scaffold trees are mainly based onchemistry-derived rules and are primarily used to map chemical space. If the generation ofscaffold trees can be guided by both chemistry and bioactivity-derived rules, chemical spaceand its related biological space can be navigated with better efficiency. Waldmann and colleaguesreported an interactive tool, named Scaffold Hunter, for intuitive hierarchical structuring, anal-ysis and visualization of complex structure, and biological activity data.98, 99 Scaffold Hunterreads data containing both chemical structure and biological activity (e.g. data from HTS).Then, the program extracts chemically meaningful scaffolds and iteratively deconstructs thoselarge scaffolds (“child” scaffolds) one ring at a time to create small scaffolds (“parent” scaffolds).Biochemical and biological activities were used as major criteria to guide hierarchical arrange-ment of parent scaffolds and children scaffolds to create a “tree” with various “branches.”Thus, the resulting tree can be associated with potency data. The method has been validatedby retrospective analysis of two large databases, PubChem and WOMBAT. The advantages ofScaffold Hunter include: (i) it investigates large chemical and biological spaces more rapidlyand efficiently; (ii) it can identify virtual (or new) scaffolds that possess bioactivity similar tothe respective child or parent scaffolds; (iii) it can simplify structurally complex compounds(e.g. natural products) to two-ring to four-ring scaffolds with retained bioactivity that are syn-thetically tractable and can be used to design new active chemotypes. Figure 3 outlines theprocess of Scaffold Hunter for identification of new active scaffolds. The seven-ring scaffold5-lipoxygenase (5-LOX) inhibitor 1 was successively deconstructed one ring at a time.98 Abranch of smaller molecules, except 5, were annotated with 5-LOX inhibitory activity in WOM-BAT. After testing for 5-LOX inhibitory activity, the three-ring scaffold 5 (IC50 = 9.5 μM) andits derivative 8 (IC50 = 3 μM) were found to be novel 5-LOX inhibitors. Although they are lessactive than the four-ring scaffold 3 (IC50 = 1.5 μM), compound 5 had higher LE values.100

Moreover, compounds 5 and 8 did not contain typical functional groups found in classical5-LOX inhibitors and represent a new scaffold for hit optimization. By a similar procedure, atetrahydroisoquinoline scaffold was identified to possess estrogen receptor α (ERα) antagonis-tic activity,98 and subsequent hit optimization led to novel ligands with a simple two-ring coreand good selectivity toward ER β (ERβ).101

4. ACTIVE SITE MAPPING AND CHARACTERIZATION BY FRAGMENT-BASEDAPPROACHES

An initial step of SBDD is the identification of hot spots in the binding pocket or active siteof the drug target. The hot spots (or consensus sites) are important regions that can bindsmall drug-like molecules and contribute substantially to the binding free energy. Therefore,identification and characterization of such hot spots is critical for rational drug design. Multiplesolvent crystal structures (MSCS) is an experimental tool to predict ligand-binding sites of targetproteins.102, 103 To use this technique, a crystalline protein is exposed to various organic solvents



Figure 3. Illustration of the procedure for the generation of the scaffold tree and brachiation-based identificationof new active scaffolds for 5-lipoxygenase inhibitors.

(smaller fragments), and then the consensus sites on the protein’s surface that are colocalizedwith multiple solvent molecules can be treated as potential ligand-binding regions. AlthoughMSCS and other experimental methods are efficacious tools for active site mapping, theygenerally require an expensive investment in equipment and resources. As an alternative, variouscomputational approaches are available to predict the ligand-binding sites of a protein.104

There are two classes of algorithms for structure-based pocket prediction: (i) geometricalgorithms and (ii) probe mapping/docking algorithms.105 For the latter, fragments are usedas molecular probes for protein surface mapping and identification of hot spots. GRID106, 107

and multiple copy simultaneous search (MCSS)108 are two well-accepted methods for active sitecharacterization. GRID calculates three-dimensional (3D) energy maps around protein bindingsites, thus highlighting favorable sites for small functional groups. MCSS randomly placesthousands of copies of small functional groups into the binding site, and the most energeticallyfavorable position of each copy is determined by energy minimization. The copies with the



Table I. Summary of Advantages and Disadvantages of Fragment-Based Computational Tools for ActiveSite Mapping

Method Advantages Disadvantages

GRID Global search of the entire proteinsurface

Require empirical parametrization andlack of water molecules in the model

CS-Map Better sampling, the ability to findsmall buried pockets and desolvation

term in the free energy calculation

No consideration for bondedinteractions and different dielectric

constants for different targetsFTMAP Fast FFT correlation approach to

efficiently reduce the computationalcosts

Lack of water molecules in the model

MCSS The most established method,reasonable use of physicochemical

potential functions andincorporation of the standard

molecular simulation framework

No consideration for the cooperativeeffects of water and locating

minimum enthalpy poses rather thanfinding hot spots

3D-RISM-basedmethod

A realistic model including thecoexistence of water and revealingthe dependence of ligand binding

modes on the ligand concentration

No consideration of protein structuralchange induced by ligand binding

Grand canon-ical MonteCarlo simu-lation

Fast and simple parameters withoutprior knowledge and calibration

Lack of complete validation and casesensitive

Barril’s method Nonparametric and applicable to anytarget class, and detecting hot spots

for both small molecules andmacromolecules

Computationally expensive and limitedsampling

lowest energies highlight “hot spots” of ligand binding. The methodology and application ofMCSS has been reviewed by Schubert et al.109 Besides GRID and MCSS, other computationalapproaches based on fragment mapping/docking and scoring are summarized in Table I.Earlier methods in this field have been reviewed,110, 111 and the following sections mainly focuson important progress in recent years.

Vajda’s group reported a fragment-based computational mapping program named CS-Map.90 CS-Map uses a three-step mapping algorithm81 that includes: (i) finding regions withfavorable electrostatics and solvation by rigid body search; (ii) refinement of free energy anddocking; and (iii) clustering, scoring, and ranking. As compared with earlier mapping methods,CS-Map performs better sampling of regions with favorable desolvation and electrostatics. Itsscoring function takes the desolvation effect into account and the positions of the dockedligands are clustered and ranked on the basis of their average free energies. More recently,the same group proposed a new algorithm named FTMAP that uses the Fourier transform(FT) correlation method for sampling protein–probe complexes in combination with a highlyaccurate energy function.80 FTMAP is more efficient than CS-Map and free to academic users.A recent validation study revealed that FTMAP could duplicate the MSCS data successfullyfor two targets of Parkinson’s disease.112 Moreover, this method can discover hot spots that arenot found in the MSCS experiments.

Imai et al. used a 3D reference interaction site model (3D-RISM) to identify the mostfavorable positions and orientations of fragment molecules on a protein surface.46 A uniquefeature of the 3D-RISM-based method is that the ligand mapping calculation is performed



within a realistic model by considering ligands and water at the same level in terms of the sitedistribution. Therefore, this method may achieve “entire ligand mapping” on a protein surfacein a real solution system. In addition, the 3D-RISM-based method can investigate the influenceof ligand concentration on the binding mode.

Some computationally expensive procedures, such as molecular dynamics simulation andMonte Carlo simulation, have also been used for fragment mapping, clustering, and ranking.Clark et al. reported a rapid and practical method to compute binding free energies of a largenumber of fragment poses on the entire protein surface.113 The method uses grand canon-ical Monte Carlo (GCMC) simulation to compute ligand–protein binding and predicts theaffinities and preferred binding poses of small molecular fragments. Thus, the fragments canbe computationally assembled into higher MW lead compounds. Barril’s group developed anew method to detect binding sites based on first-principles molecular simulations.47 More-over, this method is able to quantify the maximal binding affinity that a ligand may achieve,and thus can efficiently measure the druggability of the target. Because the method is nottrained on a dataset, it is applicable to any target class. Although it is computationally de-manding, it provides very detailed information about the interaction preferences of the bindingsites.

5. FRAGMENT DOCKING AND VIRTUAL FRAGMENT SCREENING

In most experimental FBDD studies, only hundreds to thousands of fragments can be screened.In contrast, at least 250,000 fragments are commercially available,114 leaving a large portionof fragment libraries untested. Because commercially available fragments are too numerousto be screened experimentally, virtual fragment screening by molecular docking seems to bea complementary approach. However, docking and scoring fragments accurately remains achallenge. First, fragments are small in size and have low MWs. During docking calcula-tions, a number of interaction sites on protein surfaces (closely related energy minima) mightbe found to accommodate the fragment, which would lead to false docking positions. Evenif fragments are placed into the correct pocket, if the binding pocket is large, it might re-sult in incorrect binding modes.115 Second, the internal degrees of freedom of fragments aregenerally less than larger compounds. It is more difficult to predict their binding pose be-cause alternative binding might yield similar docking scores or calculated binding energies.Third, fragments are always weak binders and current scoring functions are not accurateenough to differentiate an active fragment among many nonactive fragments, because mostof the scoring functions have been developed and optimized on the basis of larger drug-likemolecules.116

Shoichet’s group reported pioneering results for fragment docking and screening ofAmpC β-lactamase inhibitors.117 A database containing 137,639 fragments were docked byDOCK3.5.54. Forty-eight top ranked fragments were subjected to an in vitro enzyme inhibi-tion assay and 23 hits with Ki values in the range of 0.7–9.2 mM were identified. For AmpCβ-lactamase, the hit rate of the in silico fragment screening (48%) was considerably higherthan both virtual screening and HTS of larger molecules. The accuracy of the docking posesof the active fragments was further investigated by solving the crystal structures of fragment–enzyme complexes. For the eight cocrystallized fragments, four fragments had good pose fidelity(RMSD range: 1.2–1.4 A) and two fragments retained most key interactions (RMSD values:2.4 A and 2.6 A). The high hit rate and docking accuracy in this case study supports thefeasibility of molecular docking to prioritize molecules from commercially available fragmentlibraries. Moreover, this study also highlighted the importance of the selection of an appropriatedocking method and scoring function. In Shoichet’s study, DOCK3.5.54 was selected to screen



fragments because its physics-based scoring function can prioritize active fragments from inac-tive ones.117 In contrast, energy-based scoring functions118 are limited in their applications forranking fragments.

Besides DOCK, several de novo drug design software, such as LUDI107 and SEED,119

also can dock fragments into the correct pocket of the active site. For example, Caflisch’sgroup developed the program DAIM,102 which can automatically decompose molecules intofragments and then select the anchor fragments for docking. DAIM uses the docking algorithmsfrom SEED103, 119 and has been validated with six different target enzymes.106 Glide120, 121 isanother efficient tool for fragment docking.122, 123 Researchers from AstraZeneca evaluatedthe performance of Glide for virtual screening of fragment inhibitors of DNA ligase andprostaglandin D2 synthase.122 The results indicated that using GlideSP with its default settingsgave the best performance and provided an enrichment that was better than random samplingand comparable to virtual screening of drug-like molecules. More recently, evaluation studiessuggested that the sampling efficacy of Glide was adequate for fragment docking, but theperformance of scoring functions required further improvement.123 Another newly reportedstudy revealed that there is no significant difference in docking performance between fragmentsand drug-like compounds.124 Better docking performance was observed for compounds withhigher LE values, mainly because they can form high-quality interactions with the target.

In 2011, Knehans et al. reported a successful example of in silico fragment screening ofdengue virus (DENV) protease inhibitors.125 A library of 149,151 fragments was obtained fromRECAP fragmentation of the drug-like molecules from the ZINC database, which were subse-quently docked into a homology model of DENV protease by AutoDock Vina.126 A total of220 top-ranked fragment hits were discovered and subsequently linked to 815 new molecules.Virtual high-throughput docking was performed again to finally select 23 candidates for bio-logical testing and two hits were proven to be active in the micromolar range. It is expected thathigher hit rates would be achieved if the accuracy of fragment docking is improved. Regardless,this computational strategy effectively mimicked the process of experimental FBDD, whichintegrates fragment library design, fragment docking, fragment linking, and biological testing.Two other studies used a similar approach for ligand design,127, 128 but the resulting moleculeshave not yet been validated by experimental studies. Fragment docking can also be used inparallel to experimental fragment screening. A recent study discovered orally active inhibitorsof Hsp90 molecular chaperone by merging structural elements of different hits derived fromparallel fragment screening and fragment docking.129

Currently, there are four kinds of strategies to improve the accuracy of fragment dock-ing. The first is to add sophisticated and intensive computational tools (e.g., MM/PBSA,MM/GBSA, and QM/MM) to the postdocking process. Gleeson et al. evaluated the per-formance of QM/MM-based models to reoptimize and rescore cross-docking poses of ninefragment-like kinase inhibitors.130 Hybrid QM/MM calculations were proven to be useful asa tool for kinase FBDD. On the contrary, Kawatkar’s results indicated that adding more com-putationally intensive procedures to Glide docking, such as MM/GBSA rescoring, did notimprove the enrichment.122 Therefore, success with these computationally expensive rescoringstrategies might be system dependent.

The second strategy to optimize the fragment docking procedure is to improve the per-formance of scoring functions. The main problem with fragment docking failures is that thescoring functions are often unable to distinguish the correct binding mode from the incor-rect ones.124 Marcou et al. found that scoring by the similarity of interaction fingerprints forposing and prioritizing either fragments or molecular scaffolds was statistically superior toconventional scoring functions.131 One possible option to overcome the limitation of scoringfunctions might be the development of fragment-specific scoring functions.123 To achieve thisgoal, the binding nature of fragments should be taken into account. For example, fragments



have less rotatable bonds, hydrogen bond acceptors and donors, and form fewer specific inter-actions than drug-like molecules. These features should be considered when estimating theirenthalpic and entropic contributions during the parameterization of currently available scoringfunctions. In addition, force field-based scoring protocols and other more advanced methodsfor evaluating protein–ligand interaction energies are likely to be good solutions for improvingfragment-specific scoring functions.124

The third strategy is to make the fragments larger when docking them into the active site.Fukunishi et al. proposed the replica generation (FSRG) method to optimize fragment docking.In the FSRG method, a set of larger molecules (replica molecules) are generated by addingside chains to the fragment. In the docking simulation, only complementarity between thesurface of the compound and protein was evaluated, whereas hydrogen-bonding and Coulombicinteractions were ignored.132 In a validation study where inhibitors of six target proteins werescreened, the FSRG method was proven to be effective in finding active fragments among thedecoy compounds.

The fourth strategy is to dock multiple fragments simultaneously. In real cases, multipleligands are always involved in the process of molecular recognition whereas most of the dockingmethods only dock one ligand at a time. Li et al. proposed the multiple ligand simultaneousdocking (MLSD) strategy that can mimic real molecular binding processes and improve thesampling of docking poses and scoring of binding free energy.133 The MLSD method hasbeen used for fragment-based discovery of novel inhibitors of signal transducer and activatorof transcription 3 (STAT3).134 A small library of drug scaffolds was simultaneously dockedinto hot spots of STAT3 by MLSD to identify optimal fragment combinations. Linking ofthe fragment hits in combination with similarity search and structural optimization led to thediscovery of two novel STAT3 inhibitors.

Moreover, preparing the protein structure carefully and choosing appropriate parametersare also important for accurate fragment docking.108 Molecular dynamics simulation is an effi-cient tool to investigate the conformational space of the target proteins and provide reasonableconformation or conformation ensembles for subsequent fragment docking. Ekonomiuk’s ex-ample indicated that using molecular dynamics snapshots of NS3 protease for fragment-baseddocking could identify two small-molecule inhibitors that could not be identified by simplyusing the X-ray structure.135

6. FROM FRAGMENTS TO LEADS: DE NOVO DRUG DESIGN

A. De novo Drug Design versus FBDD

Experimental FBDD uses sensitive biophysical techniques to identify low-affinity fragmentshits.4 Then fragment hits are evolved into leads or candidates by various structure-based designstrategies.15 In contrast, de novo drug design can be seen as “completely” virtual FBDD becausethey have similar concepts and objectives. Both approaches start from small fragments (buildingblocks) and aim to convert them into drug-like compounds with novel chemotypes and desiredpharmacological properties. De novo design was approximately first introduced in 1989,136 andwas regarded as a complementary approach to HTS and FBDD. In principle, de novo designis cost and time effective, and can explore larger chemical space. Although more than 30 denovo design tools have been developed,39, 62 the success of de novo design in lead discoveryand optimization lags far behind that of experimental FBDD. With respect to more than tenclinical candidates identified by experimental FBDD,6 de novo drug design rarely generatesnovel molecules with nanomolar activity. The main problems of de novo design include (i) lowefficiency in the sampling of the chemical space (or drug-like space); (ii) little consideration for



synthetic feasibility when constructing new structures; (iii) poor accuracy of scoring functionsto predict the binding free energy of the designed compounds; and (iv) unreasonable ADME/Tproperties.39

The success of experimental FBDD greatly motivates the improvement of de novo design.First, the core idea of de novo drug design is very similar to that of experimental FBDD. De-veloping new methods that can overcome the limitations of traditional de novo design methodsshould help to find high quality leads efficiently. Second, de novo design could also com-plement current experimental FBDD methods by providing potential solutions and reducingexperimental resources. Third, de novo design tools are helpful to address current challengesof experimental FBDD. For instance, it is difficult to experimentally investigate the linkers thatenable fragment hits to maintain their binding mode. In this case, de novo design tools can usesimple geometric descriptors to suggest reasonable linkers and predict the binding mode of theresulting molecules by docking and scoring.

The starting point for de novo design can be atoms or fragments. It is noted that allthe atom-based tools were developed two decades ago.39 Atom-based approaches have theadvantage that they can systematically search both the chemical space and structural diversity.However, these tools always suggest a huge number of potential solutions and the resultingmolecules are often problematic in terms of chemical stability, synthetic possibility, and druglikeness. These limitations can be overcome by fragment-based approaches because the searchspace can be significantly reduced. The synthetic accessibility and drug likeness of the designedligand can be more readily captured. The typical components of a computational fragment-based de novo design process include a fragment library, a compound build-up scheme, ascoring function, and an optimization procedure. Table II summarizes the key features of newde novo design methods reported from 2006 to 2011. In the following sections, we will presentrecent progress made in the methodologies and applications of fragment-based de novo designwith a focus on current solutions to address the problems of de novo design.

B. Strategies to Assemble and Optimize Fragments

There are two major strategies to assemble fragments into novel molecules: fragment-basedgrowing and/or linking strategies and fragment hybridization strategies. It is estimated thatthe search space of drug-like molecules is about 1060 compounds, which makes it difficult toevaluate all solutions in a reasonable computational time. Thus, the realistic function of denovo drug design is to efficiently find “good” solutions rather than find the best solution. Theoptimization algorithms can improve the efficiency of sampling the huge chemical space of drug-like molecules and guide the search to appropriate regions where candidate molecules can befound. “Natural Computing” algorithms including particle swarm optimization (PSO), evolu-tionary algorithms (e.g. genetic algorithm, evolution strategy),137 and ant cloning optimization(ACO)138 have been widely used in de novo drug design.139 Such nature-inspired techniques areuseful for searching very large chemical space to converge on chemically meaningful optima.

1. Fragment-Based Growing and Linking StrategiesFor the growth approaches, predefined conformation between a seed (fixed scaffold) and thebinding pocket of the receptor is necessary. The binding complex can be obtained from molecu-lar docking or experimental techniques. Then, the seed grows fragment by fragment to comple-ment the active site geometrically and energetically. At each step of fragment growth, scoringfunctions are used to accept or reject the modifications. Early methods of this type include:SmoG,140, 141 GrowMol,142 GroupBuild,143 SPROUT,144 and GROW.145



Table II. Important Features of the De Novo Design Methods Reported from 2006 to 2011.

Methods Building blocks Assembly rules Search procedure Fitness functions

Fragmenthopping

Five general-purposelibraries: the basic

fragment library, thebioisostere library,

the rules formetabolic stability,

the toxicophorelibrary, and the side

chain library

Fragment linkingby LUDI

Mappingfragments onthe minimal

pharmacophoricelements

Consensus dockingscoring, ADMET

filters

AutoGrow Filtering the ZINCdatabase by atom

numbers

Fragment-basedgrowing

strategies

Evolutionaryalgorithms

AutoDock scores

COLIBREE Pseudoretrosyntheticfragmentation of sixcompound libraries

A fixed build-upscheme of

scaffold, linkers,and building

blocks

A discrete versionof particle

swarmoptimization

(PSO)

Using CATStopological

pharmacophoresimilarity to

reference ligandsas fitness function

FlexNovo Large fragment spaceswith up to several

thousand fragments

Sequential growthstrategy,considersynthetic

possibility bywell-defined

connection rules

Incrementalconstruction

algorithmwithin FlexX

Scoring functionsfrom FlexX,

physicochemicalproperty filters,pose geometry,

and diversity filter

Flux Virtual retrosynthesisof the COBRA

dataset by RECAP

A restricted set ofreactionschemes

Evolutionaryalgorithms

Ligand-basedsimilarity scoring

using referencecompound

FOG Collection ofconnectivity statistics

for fragments ofinterest from a

database of smallmolecules

The sequentialgrowth of small

moleculesconstrained bythe transition

probabilities ofthe growthfragment

Statisticallybiasing thegrowth of

molecules withdesired features

A linear scoringalgorithm(TopClass)

Fragmentshuffling

Alignment of theprotein–ligand

complexes, ligandfragmentation, and

calculation offragment score

Incrementalconstruction ofnovel ligands

A tree searchalgorithm

QXP scorestogether with theoverall fragment

scores

GANDI Fragment library frommolinspiration

cheminformatics

Linkingpredocked

fragments bySEED

A geneticalgorithm and a

tabu search

A linearcombination of

three scoringfunctions

including forcefield energy and

2D or 3Dsimilarity



Table II. Continued


MED-Hybridise UsingMED-SuMo-Fragmentor to

define 3Dprotein-fragment

patterns calledMED-Portion

Recombiningchemical

moieties fromMED-Portions

into putativeligand molecules

QueryMED-SuMo

database

MED-SuMo hitdescriptors,

bioinformaticdescriptors, andphysicochemical

properties;target-specific

filters forhybridizedmolecules

MEGA A substructuremining toolincludingfrequentsubgraph

mining andRECAP rule

Growing strategy Combinesmultiobjectiveevolutionary

techniques withgraph theory

Binding affinityscorers, molecularsimilarity scorers,

and chemicalstructure scorers

SQUIRRELnovo Multiconformerdatabase (17,934

fragments) byRECAP-basedfragmentation

of COBRAcollection

Bioisostericreplacements

Alignment offragments to a

referencecompound by

graph matchingalgorithm

LIQUID “fuzzy”pharmacophore

function

Hecht’s method Fragmentation ofliterature

libraries intoscaffolds and

R-groupfragments

Fragment linkingand docking

Evolved fragmentassembly

Binding affinityscore and

artificial neuralnetwork

Recore Fragmentation ofCSD database

by RECAP rulesand filtered byvarious criteria

Scaffoldreplacement

R-tree index andk-nearest-neighbor

search

Geometric ranking

EAISFD A fragmentlibrary with over1300 fragmentsextracted from

MDDR

Scaffold hoppingor substructure

optimization

Evolutionaryalgorithm

Surflex-Dock score

LigBuilder 2.0 Extraction fromWDI database

Growing andlinking

Genetic algorithm Binding affinityprediction,

physicochemicalpropertiesevaluation,

lock–key matchevaluation,

synthesizabilityprediction



Table II. Continued


NovoFLAP A fragmentlibrary with over1300 fragmentsextracted from

MDDR

Modification orhopping of the

starting structure

Evolutionaryalgorithm

(EA-Inventor)

Ligand-basedscoring function:

flexible ligandalignment

protocol (FLAP)PROTOBUILD 594 fragments

derived fromdecomposition

of ThomsonPharmadatabase

Fragment growthin the binding siteaccording to a set

of fragment–fragment

interconnectivityrules

Genetic algorithm PROTOSCOREthat includes

improved termsfor the estimation

of entropy plusterms for various

nonbondedinteractions

PhDD Eight types offragmentdatabases

(correspond toeight popular

pharmacophorefeatures) by

fragmentationof MDDR and

CMC

Linking fragmentsthat are fitted topharmacophore

model

Alignment offragments to thepharmacophore

hypothesis

Fitness score to thepharmacophore

hypothesis,assessment of

drug likeness, andsynthetic

accessibility

Fragment-based growing strategies are computationally efficient and the generatedmolecules always have good chemical diversity. Moreover, the growth methods always com-bine docking software for binding pose prediction and scoring. For example, the incrementalconstruction algorithm of docking software FlexX146 was used to develop the new de novo de-sign algorithm FlexNovo.147 FlexNovo works with the fragments directly and deals with largefragment spaces (several thousand fragments). Moreover, various filters including physicochem-ical properties, diversity, and placement geometry have been incorporated into the fragmentbuild-up process. More recently, FlexNovo was implemented into a comprehensive projectnamed NovoBench.148 The NovoBench project integrates several tools, such as generationof fragment space (Colibri149 and FragView), structure-based (FlexNovo147), property-based(FragEnum150), and ligand-based (Ftrees-FS64) search algorithms to meet the various demandsof de novo design.

In many growing strategies, the binding mode or pose of the “seed” fragment is assumedto be fixed upon fragment growth. However, this assumption is often invalid in many cases.Durrant et al. developed AutoGrow that combines elements of fragment-growing, docking,and evolutionary algorithm.151 In AutoGrow, each generated new compound during fragmentaddition is dynamically redocked into the binding pocket by AutoDock152 and generates newposes for each molecule. An evolutionary algorithm is used to explore new chemical space byevaluating the docking scores of every population member, and select the best molecule forsubsequent generation.

Another method, FOG, also grows molecules by adding fragments to a nascent molecule.153

The novelty of FOG lies in that the growth of molecules depends on the frequency of specificfragment–fragment connections by mining a specific molecular database. In addition, FOG canbe trained to grow new molecules with chemical and topological features similar to a desired



class of compounds (e.g. natural products and drugs) by the Topology Classifier (TopClass)algorithm.

Fragment-based linking strategies map fragments onto the key regions in the bindingpocket to determine various energetically favorable positions and then link them together tobuild new molecules. The above-mentioned active site mapping approaches, such as GRIDand MCSS, are commonly used to position seed fragments or functional groups to the correctlocations in the binding pocket. The “link” concept has been widely used in a number of denovo design methods including CONCERTS,154 LUDI,155, 156 CAVEAT,157 NEWLEAD,158

DLD,159 BUILDER,160 and SKELGEN.161 These methods can search large chemical spaceusing a relatively small number of initial fragments. The resulting molecules are expected tobind more tightly with the target than the individual fragment. However, fragment linking mightlead to a change of the overall conformation of the generated molecules, and key interactionsbetween the initial fragment and target might be lost. Thus, redocking the new ligands mightbe necessary in the postprocessing step.

GANDI is a new de novo design tool for automatically linking predocked fragmentswith a user-defined fragment library.162 GANDI uses SEED163 for fragment docking andits optimization procedure combines a genetic algorithm and a tabu search. An importantfeature of GANDI is its multiobjective evolutionary optimization strategy that simultaneouslyoptimizes the force field energy and a 3D-overlap term to known binding modes or a 2D-similarity term to known inhibitors. Thus, GANDI can be both structure-based and ligand-based according to the user’s need. In addition, GANDI is free to academic users, whichprovides more opportunities to validate the method.

Ji et al. proposed “fragment hopping” as a new fragment-based tool for de novo inhibitordesign.164, 165 Fragment hopping is a pharmacophore-driven strategy focusing on isozyme se-lectivity and ligand diversity. The derivation of the minimal pharmacophoric element for eachpharmacophore is the key point of this approach. The minimal pharmacophoric element canbe an atom, a cluster of atoms, a virtual graph, or vector(s), which can be derived from acombinatorial application of different active site analysis and pharmacophore identificationmethods. The novelty of minimal pharmacophoric elements lies in that they can map an impor-tant interaction pattern between a ligand and hot spots for both isozyme selectivity and ligandbinding based on a priori knowledge and experience. Five fragment libraries are implementedwithin fragment hopping. The basic fragment library and the bioisostere library are queriedto generate a focused fragment library with diverse structures that can match the requirementsof the minimal pharmacophoric elements. Then, the focused fragment library is filtered by therules for metabolic stability and the toxicophore library. The binding positions of the resultingfragments to each pharmacophore are searched by LUDI and the MCSS program. Finally, thedesired molecules are generated by linking these fragments using the side chain library. Theevaluation process includes docking, consensus scoring, and ADMET filters. Moreover, thisnew de novo design methodology is an open and interactive system according to the medicinalchemists’ requirements for a specific research project. The application of fragment hopping tode novo design of highly potent and selective neuronal nitric oxide synthase (nNOS) inhibitorswill be introduced in the section of case studies.

More recently, Lai’s group developed LigBuilder 2.0,166 which is an improved version oftheir previously reported method LigBuilder 1.0.167 LigBuilder uses a genetic algorithm toconstruct ligands iteratively by fragment linking or growing. Compared with LigBuilder 1.0,the new version takes synthetic accessibility into account by an embedded chemical reactiondatabase and a retrosynthesis analyzer (SYLVIA).168 Moreover, an accurate cavity detectionprogram (Cavity 1.0) and the “Drug Space Exploring Algorithm” was incorporated into thedesign process. Various filters including binding affinity evaluation, physicochemical propertiesevaluation, and lock-key match evaluation are used to find the best molecules.



2. Molecular Hybridization and Scaffold HoppingMolecular hybridization is a common design strategy in medicinal chemistry.169, 170 Thecore idea of molecular hybridization is to combine pharmacophoric fragments from differ-ent bioactive compounds to generate a new hybrid molecule with improved biological orphysicochemical properties. This concept has been used to develop new de novo design algo-rithms. BREED is the first automated method that produces novel molecules from structuresof different known ligands targeting a common receptor.171 This method takes advantageof the structural information of known ligands–target complexes, aligns the 3D coordi-nates of two ligands, and recombines fragments at overlapping bonds to generate hybridmolecules.

Inspired by the idea of the BREED method, Wang’s group developed an automatic methodnamed automatic tailoring and transplanting (AutoT&T) that can effectively utilize the resultsof virtual screening in fragment-based lead optimization.172 AutoT&T identifies suitable frag-ments from virtual screening hits and then transplants them onto a predefined lead compoundto generate new ligand molecules with improved binding affinities. As compared with the con-ventional de novo design methods, AutoT&T has several advantages. First, it detects fragmentsdirectly from other organic molecules and does not rely on a predefined building block library.The input molecule databases can be flexible with no limitations in terms of sizes or types.Second, AutoT&T is more efficient and does not have the problem of combinatorial explosion.It performs structural transplantation on the basis of the matched bonds between the leadcompound and each given steak molecule without adopting a sequential build-up approach.Third, synthetic feasibilities and drug-likeness properties are taken into account during theinvention of new molecules.

The idea of “mix and match” in BREED was also used to develop several new algorithms,such as MED-Hybridise and fragment shuffling. Fragment shuffling differs from BREED inthe way that it is able to hybridize multiple ligands within one iteration step and includesfragment scores to guide the incremental construction of the new ligands.173 MED-Hybridiseis a computational drug design toolkit at PDB scale that combines the local similarity ofprotein surfaces and a fragment-based approach.174 MED-Hybridise takes advantage of therich structural information of target–ligand complexes in the PDB database. An important stepin MED-Hybridise is to define the MED-Portion, which is the 3D protein-fragment patternsobtained from mining all available protein–ligand crystal complexes within a library of smallmolecules. For any binding surface query, matched MED-Portions can be retrieved usingMED-SuMo175, 176 to superimpose similar protein interaction surfaces. The resulting MED-Portion chemical moieties are collected and used to generate new 3D hybrid molecules. In aretrospective validation study, MED-Hybridise could successfully retrieve scaffolds of knownactive compounds for a GPCR target (β2-adrenergic receptor) and a protein kinase target(vascular endothelial growth factor receptor 2, VEGFR-2).

The concept of scaffold hopping177 is similar to that of molecular hybridization. Thecore idea of scaffold hopping is to replace a central element of the molecular scaffold bya new molecular fragment. There are several computational tools for scaffold hopping.178

Recore is a fast and effective approach for scaffold replacement that uses 3D fragmentsas queries and can search pharmacophore-type features.179 Moreover, Recore incorporatesk-nearest-neighbor searches and a voting system to enable the exploration of large searchspaces.

3. Ligand-Based MethodsAlthough the majority of de novo design tools are structure-based methods, they also faceseveral major challenges such as the accuracy of the scoring functions and the flexibility of the



receptor. Moreover, it remains a challenge to solve the crystal structures of many membrane-bound drug targets (e.g. GPCRs). In this case, ligand-based methods provide an alternativefor de novo drug design. Such approaches do not rely on the 3D structure of the drug targetand instead, their design process is guided by maximizing the similarity between the generatedmolecules and the known active compounds.

Ligand-based methods are an emerging hot area in recent years. Schneider’s group hasproposed three promising approaches (i.e. COLIBREE, Flux, and SQUIRREL).62, 180 COL-IBREE uses a fixed build-up strategy by adding various building blocks and linkers to apredefined molecular scaffold to generate a focused combinatorial library.181 The optimiza-tion procedure is guided by a stochastic optimization algorithm, PSO,180 with CATS topo-logical pharmacophore similarity177 to reference ligands as a scoring function. Flux assem-bles molecules by a restricted set of reaction schemes and the chemical synthesis of theconstructed molecules is eased using molecular building blocks obtained from the RECAPprinciple.182, 183 A stochastic search algorithm is implemented in Flux with similarity-baseddescriptors and metrics as fitness functions. SQUIRREL is a new algorithm to compare bothmolecular shape and potential pharmacophore features.184 It was used to develop a ligand-based de novo design tool that can suggest bioisosteric replacement groups for a referencecompound.185

Tripos’s EA-Inventor is a generic structure invention engine based on an evolutionaryalgorithm.186 It works on the connection tables of an initial population of structures and theevolutionary process of structure invention. It can be driven by any user-defined scoring func-tion (binding affinity, pharmacophores, similarity, or other desired properties). EA-Inventorhas been used in several structure-based or ligand-based de novo design approaches.187–190 Liuet al. reported a structure-based method named EAISFD,189 which combines EA-Inventor forstructure evolution with Surflex-Dock191, 192 for docking and scoring. EAISFD introduced the“Tagged Fragment” (TF) strategy for the multiobjective build-up process. TF can be either afragment (substructure) of the ligand or a new fragment attached to the ligand that serves to an-chor key binding interactions. Thus, the TF strategies can be used for partial or full drug design,such as scaffold hopping, substructure optimization, and structure extension. Now, EAISFDhas been developed into a commercialized product, namely Muse (http://www.tripos.com),for de novo drug design. More recently, researchers from Tripos reported NovoFLAP as aligand-based approach that combines EA-Inventor with a powerful scoring function FlexibleLigand Alignment Protocol (FLAP). FLAP uses both molecular shape and pharmacophoricfeatures in a multiconformational context.187

Pharmacophore hypothesis can also be useful in de novo drug design. Fragments that fit indifferent parts of a pharmacophore model can be linked together by various spacers to gener-ate novel structures. NEWLEAD is the first pharmacophore-based de novo design method.158

However, it can only process pharmacophoric functional groups rather than abstract chemi-cal features such as hydrogen bond donors and acceptors, and hydrophobic features. Yang’sgroup made important improvements for pharmacophore-based de novo design. Their newmethod, PhDD, is able to work with abstract pharmacophore models and be implemented withcomprehensive evaluators including drug likeness, bioactivity, and synthetic accessibility.193 Aresearch group from Eli Lilly reported a fragment-based method for the de novo design ofkinase inhibitors.194 Fragmentation of existing kinase inhibitors was used to generate buildingblocks that were subsequently recombined to create “de novo” chemical libraries. The librarieswere driven by a general kinase pharmacophore model and a support vector machine basedmethod (SVMFP)195 was used to predict combinations of fragments. The overall hit rate of thepharmacophore-driven de novo library was very high (92%), which highlights the superiorityof this fragment-based strategy over virtual screening or structure-based minor modificationsof existing inhibitors.



C. Scoring Functions and Multiobjective Optimization

Scoring functions are crucial to guide the optimization process and to evaluate the bindingaffinity or chemical similarity of the designed molecules. During the process of sampling chem-ical space, a huge number of iterations should be evaluated by a scoring function. Thus, scoringfunctions are required to be fast, but this feature comprises their accuracy. Scoring functionsfrom molecular docking are popular in structure-based de novo methods.196 These scorers areadept at discriminating between inactive and active compounds, but their ability to rank theligands with similar chemotypes is relatively poor. Therefore, the application of docking-basedscoring functions in de novo design is less successful than in virtual screening. On the otherhand, comprehensive physics-based approaches, such as MM-GBSA/PBSA,197, 198 free-energyperturbation,199, 200 single-step perturbation,201 GCMC simulations,113 and thermodynamicintegration,202 can yield very accurate binding free energies. However, these methods requirehigh computational expenses, and thus are not suitable for the search process of de novo design.Even so, they are very helpful to improve the success rate of de novo design by re-evaluating thefinal molecules after postprocessing and identifying the best one for chemical synthesis and bi-ological testing. Nowadays, the availability of computational resources (e.g. cloud computing)is increasing dramatically, which will enable a wider use of physics-based scoring functions inde novo design.

Computational intelligence, such as machine learning approaches, has been used in drugdesign for the automatic selection of important features and the optimization of models.203 Incombination with traditional scoring functions, computational intelligence tools can quicklyand efficiently search diversity space for good solutions. Hecht’s method uses an evolved frag-ment assembly algorithm for directed searches of novel leads.204 The novelty of the methodlies in that it uses a computational intelligence screening tool for compound selection. Thescreening tool integrates evolved artificial neural nets, docking software as well as QSAR andQSPR models.

Early de novo design approaches have been created to satisfy a single objective and most ofthem have focused on the interaction scores with the binding pocket. However, these methodsignore the multiobjective nature of drug discovery and development. Thus, the application ofmultiobjective optimization strategies in design workflow is beneficial to improve the drug-likebehavior of the generated molecules. Moreover, multiobjective optimization methods can avoidlocal optima and dead ends corresponding to a single objective and lead to a more efficientsearch process. MEGA is a new de novo design algorithm for multiobjective optimizationthat combines graph theory with evolutionary techniques to perform an efficient global searchfor promising solutions.205 Three kinds of fitness functions, namely binding affinity scorers,molecular similarity scorers, and chemical structure scorers, are used to guide the optimizationprocess. Therefore, structurally diverse molecules with good binding energy and drug-likeproperties can be designed by MEGA.

D. Synthetic Accessibility and Drug Likeness

Synthetic accessibility is one of the key issues that remain to be addressed in de novo design.Two types of approaches have been reported to improve the synthetic accessibility of computer-designed structures. The first approach uses connection rules and synthesizable building blocksto construct new molecules. Synthesizable building blocks can be obtained either from decom-position of compound databases by virtual retrosynthesis rules or from commercially availablecompounds. The connection rules are mainly derived from organic synthesis reactions.206, 207 Asmentioned above, RECAP59 rules have been commonly used to disassemble and reconstructsynthetically feasible molecules (e.g. TOPAS206 and Flux). The advantage of using RECAP



rules in de novo design is that the chemical environment around a new bond is similar tothat in drug-like and synthesizable compounds. Compounds created from such rules are moreapproachable. The limitations of RECAP lie in that they are crude abstractions of actualchemical reactions and only cover a small number of types. Moreover, compounds constructedfrom RECAP rules are only potentially synthesizable, and no scores and synthetic routes canbe suggested. Other methods, such as SYNOPSIS,207 use known chemical reactions to formbonds between readily available building blocks. A total of 70 selected organic reactions wereimplemented in SYNOPSIS and the building blocks are taken from the ACD208 database.

The second approach is amenable to the first one and uses additional scoring functionsto evaluate the synthesizability of the generated candidates in the postprocessing step. For ex-ample, FOG generates synthetically tractable molecules by use of the software168 SYLVIA.153

Other computer-aided organic synthesis design methods, such as Route Designer,209 can sug-gest synthetic routes for experimental validation studies of de novo design. More recently, newcheminformatics tools (e.g. reaction vectors210 and Reaction-MQL211) for the representationand encoding of organic reactions should contribute to the de novo design of molecules withgood synthetic accessibility. A new algorithm named DOGS is under development by Schnei-der’s group.62, 212 DOGS is based on a series of known reactions selected from the literature andcan propose at least one possible synthetic route for each designed compound. Other methodsfor synthetic evaluation are reviewed by Kutchukian et al.213

Drug likeness is another important constraint in de novo design. As described above, theconsideration of drug likeness has been incorporated into every stage of de novo drug design.The fragment libraries for molecular invention are often obtained from the decomposition of adrug or drug-like database, which is based on the assumption that molecules built from drug-likebuilding blocks are more likely to possess drug-like properties. An efficient search of drug-likechemical space is also necessary to invent high-quality molecules. In the postprocessing stage,Lipinski’s rule of five214 is used as a popular filter for prioritizing candidates before synthesisand biological evaluation. Moreover, recent progress in computational ADMET predictionstrategies215 can also be applied to de novo drug design, which is helpful to re-evaluate thecandidate molecules in a cost-effective manner.

E. Recent Examples of Fragment-Based De Novo Design

Although de novo design tools are far from perfect and rarely generate ligands with nanomolaractivity, they can be viewed as an “idea generator” to provide novel chemotypes for medici-nal chemists.196 Integration of de novo design software and expertise knowledge of medicinalchemists is likely still necessary to find high-quality hits or leads. Table III summarizes re-cent examples of de novo design.216–233 In combination with our own work in this field, foursuccessful examples will be discussed in detail.

1. Case 1: De Novo Design of Selective nNOS InhibitorsNOS represent a family of enzymes that produces nitric oxide (NO). There are three isozymicforms of NOS including endothelial (eNOS), macrophage or inducible (iNOS), and neuronal(nNOS) isozymes. Among them, nNOS is an important target for various neurodegenerativedisorders.234 However, structure-based design of isoform-selective inhibitors is a difficult andchallenging task because the active sites are nearly identical for all three NOS isoforms.235–238

Ji et al. reported successful examples for the de novo design of selective nNOS inhibitors on thebasis of the minimal pharmacophoric elements and fragment hopping.164 The active site of NOSwas investigated by two different methods (i.e. GRID and MCSS) and the obtained informationcombined with previous SAR results led to the generation of the minimal pharmacophoric



Table III. Selected Examples of Fragment-Based De Novo Design from 2006 to 2011



Table III. Continued



Table III. Continued



Figure 4. A Minimal pharmacophoric elements for selective nNOS inhibitor design. B Binding mode of inhibitor9 with nNOS and hydrogen bonds are displayed as red dash lines. C Chemical structures of two novel nNOSinhibitors derived by fragment-based de novo design. The structural information is obtained from the ProteinDatabank (PDB code: 3B3N).

elements for nNOS. The minimal pharmacophoric elements identified for selective nNOSinhibitor design included an amidino group, four nitrogen atoms, and two hydrophobic (orsteric) groups (Fig. 4A). An amidino group is positioned close to E592 of nNOS to formcharge–charge and hydrogen-bonding interactions. A sp3-hybridized nitrogen cation is placedclose to the selective region defined by D597, while the other three nitrogen atoms are near to theheme propionate to form charge–charge interactions and hydrogen bonds. The regions wherehydrophobic and/or steric interactions play important roles are the positions closest to D597and the heme propionate. Using a design process of fragment hopping, a focused fragmentlibrary was generated to match the minimal pharmacophoric elements and the fragments werelinked by LUDI to build new molecules.

After evaluation, compound 9 (Fig. 4C) and its analogues were subjected to chemicalsynthesis and inhibitory activity testing. Compound 9 revealed potent inhibitory activity towardnNOS (Ki = 388 nM). Moreover, it was also a highly selective nNOS inhibitor with 1100-foldand 150-fold selectivity over eNOS and iNOS, respectively. The crystal structure of nNOS incomplex with compound 9 indicated that only the (3′S, 4′S)-isomer was bound and its bindingconformation was similar to that obtained from docking. The nitrogen atom of the pyrrolidinering interacted with the region for selectivity, which is lined with residues nNOS Asp597/eNOSAsn368 (Fig. 4B). The nitrogen atom next to the pyrrolidine ring formed a hydrogen-bondinginteraction with one of the heme propionate groups and the terminal amino group formedcharge–charge and hydrogen-bonding interactions with the heme propionate of the pyrrole Dring and one structural water.



Although compound 9 is a selective inhibitor of nNOS, it is too hydrophilic to cross theblood-brain barrier. Subsequent optimization studies were focused on increasing its inhibitoryactivity and lipophilicity.165 Two hydrophobic/steric pockets of nNOS were identified by frag-ment hopping, which were used to generate minimal pharmacophoric elements for better ligandpotency as well as better physicochemical and pharmacokinetic profiles. After introducing amethyl group on the 2-aminopyridine ring and a substituted phenylethyl group on the terminalamino group, compound 9 was evolved into 10, which formed stronger hydrophobic inter-actions with these two pockets. Compound 10 showed higher inhibitory potency for nNOScompared to 9. The Ki value for the racemic mixture was 14 nM, and its nNOS selectivityover eNOS was 2000-fold. Further studies revealed that (3′R, 4′R)-10 showed better activityand selectivity than the other isomers.239 The Ki value for (3′R, 4′R)-10 with nNOS was 5.3nM and its selectivity of nNOS over eNOS and iNOS was more than 3800-fold and 700-fold,respectively. Moreover, a racemic mixture of compound 10 also showed good in vivo efficacy ina rabbit model for cerebral palsy.240 This study also validated the ability of fragment hoppingto achieve better inhibitory potency and isozyme selectivity, which are important aspects forlead optimization.

2. Case 2: De Novo Design of Small Molecule Inhibitors of Cyclophilin ACyclophilin A (CypA) plays an important role in numerous biological processes.241–243 Smallmolecule CypA inhibitors can be used for developing immunosuppressive, antitumor, and car-diovascular agents. Li’s group found that the amide fragment is an important pharmacophorefor CypA inhibitors. The amide group functions as the key linker between two terminal frag-ments and interacts with the “saddle” between two sub-binding pockets of CypA. On the otherhand, the urea fragment has also been used as a linker for CypA inhibitors.244 By fusing amideand urea, a new linker, acylurea, was designed as the seed for de novo drug design (Fig. 5A).245

Using LigBuilder 2.0,166 the seed structure grew from both ends to fit the shape and propertiesof the two sub-binding pockets in CypA (Fig. 5B). The generated molecules containing differ-ent R1 and R2 fragments were ranked according to the scores of binding affinity, biologicalavailability, shape complementarity, and synthesizability. Finally, compound 11 was chosen forchemical synthesis and was proven to be a highly potent CypA inhibitor (IC50 = 31.6 ± 2.0nM). Subsequent SAR studies revealed that the placement of the 9H-fluorene ring by otheraromatic groups was not tolerated. The optimization of the 2,6-dihyroxy substituents yieldeda more potent inhibitor (compound 12, IC50 = 1.52 ± 0.1 nM). Figure 5C depicts the bind-ing mode of compound 11 with CypA. The planar fluorene ring and 2,6-disubstituted phenylmoiety fit two hydrophobic areas in site A and B, respectively (Fig. 5C). The acylurea linkerformed seven hydrogen bonds with residues Arg55, Gln63, and Asn102 in the saddle site. Inparticular, compound 12 represents the most potent CypA inhibitor reported to date.

3. Case 3: Optimization of 5-HT1B Receptor Antagonists by a Ligand-Based De Novo DesignMethodBesides generating entirely novel molecules, de novo design can also be used to modify existingstructures. A research group from AstraZeneca reported a successful story of lead hoppingand optimization of 5-HT1B receptor antagonists using the ligand-based de novo design toolNovoFLAP (Fig. 6).246 Based on the chroman template 13, a library was generated and rankedby NovoFLAP by focusing on the replacement of the basic piperazine moiety. Pyrazole-basedcompound 14 was initially selected because of its potential ability to form hydrogen-bondinginteractions with an aspartic acid residue of the 5-HT1B receptor. After taking chemical ac-cessibility into account, a small library of pyrazole–quinolone derivatives was synthesized.Compound 15 was the most potent ligand toward the 5-HT1B receptor (Ki = 9.3 nM), but it



Figure 5. A The process of fragment-based de novo design of novel CypA inhibitors. B The general bindingpattern of CypA inhibitors. C The binding mode of novel inhibitor 11 with CypA.

Figure 6. Optimization of 5-HT1B receptor antagonists by the ligand-based de novo design tool NovoFLAP.



was found to be a partial agonist. Subsequent optimization efforts led to the discovery of com-pound 16 as a full antagonist, which showed improved activity with a highly selective profile(Ki = 0.43 nM). Moreover, compound 16 showed a dose-dependent reversal of agonist-inducedhypothermia in an in vivo model. The new series obtained from de novo design represent thefirst nonbasic (without the ubiquitous piperazine moiety) antagonists of the 5-HT1B receptor.This example also reveals the promise of de novo design for lead optimization.

4. Case 4: Fragment-Based De Novo Design and Optimization of CYP51 InhibitorsLanosterol 14α-demethylase (CYP51) is an important antifungal target. Triazole antifungaldrugs (e.g. fluconazole and voriconazole) are classical CYP51 inhibitors and are used as first-line antifungal agents. 3D models of fungal CYP51s have been constructed by our group usingcomparative homology modeling.247–249 On the basis of the CYP51 models, computationalfragment-based methods have been applied to optimize triazole antifungal agents and designnovel CYP51 inhibitors.250 With an aim to find highly potent and broad spectrum triazoleantifungal agents, our group proposed a fragment-based pharmacophore model to improve theefficiency of azole optimization (Fig. 7).251 In this model, four kinds of functional fragmentsincluding linker, aromatic or alkyl group, hydrogen bond acceptor, and hydrophobic groupwere discovered as important for CYP51 binding and antifungal activity. Various functionalfragments have been designed and combined to fit the model (Fig. 7), and most of the re-sulting compounds showed good antifungal activity against clinically important pathogenicfungi.252–258 Several new triazole derivatives showed better activity than itraconazole and flu-conazole with minimum inhibitory concentration (MIC) values ranging from 0.25 μg/mL to0.001 μg/mL. Among the highly active triazoles, a promising candidate named iodiconazolewas selected for drug development and is currently in a phase III clinical trial.259–261 Our grouphas used fragment-based methods for the de novo design of novel CYP51 inhibitors that areexpected to overcome the cross-resistance and hepatic toxicity of triazole antifungal agents.262

The active site of CYP51 was explored by MCSS and the key regions essential for inhibitorbinding were identified. Various MCSS minima were linked to a benzopyran scaffold by LUDIto afford a series of new inhibitors. Compound 17 had an IC50 value of 35.21 μM toward Can-dida albicans CYP51 (Fig. 8) and represents a good lead for further optimization. Extensivescaffold hopping studies have been performed to find a better core fragment (Fig. 8).263–266

The aminotetralin scaffold and benzoimidazole scaffold were also found to be favorable for theantifungal activity.263, 266 Because these novel inhibitors do not coordinate with the heme andinteract with CYP51 mainly through nonbonding interactions, they are promising leads for thedevelopment of selective antifungal agents with better safety profiles.

7. CONCLUSIONS

Considerable developments have been made in the methodology and application of FBDD overthe past 15 years. Despite these successes, experimental FBDD requires high-quality protein,highly soluble fragments, expensive detection equipment, and specific expertise, which limitits applications to selected laboratories and biological targets. In this context, cheminformaticsand computational chemistry can be used as alternative approaches. Computational approachescan play a synergistic role in each step of experimental FBDD by maximizing its performance.Computational FBDD can also be used independently in generation and optimization of leads.In particular, fragment docking and de novo design methods have been widely used, but thereare limited good examples. Even so, rapid progress has been made in these fields and successfulexamples are emerging.



Figure 7. Fragment-based optimization of triazole antifungal agents.

In order to improve the efficiency and applications of FBDD, future research should befocused on methodology development and integration. For experimental FBDD, the method-ologies are required to be continually developed, and includes the construction of improvedfragment libraries that cover larger drug-like space, the development of higher throughput, andmore sensitive biophysical methods for fragment screening and the improvement of strategiesfor evolving fragments into leads. For computational FBDD, it is of key importance to developnew methods that can accurately predict the binding pose of fragments and estimate the bind-ing free energy (both entropy and enthalpy predictions). Although such developments are alsochallenging for current computer-aided drug design methods, the accuracy of computationalFBDD needs to be at least comparable to those of drug-like molecules. De novo design has



Figure 8. De novo design of novel CYP51 inhibitors by MCSS/LUDI and the optimization of the core fragmentby scaffold hopping.

great potential for both inventing novel structures and optimizing existing leads. Inspired by thesuccess of FBDD, rapid progress has been made in the methodologies of de novo design in thelast 5 years. Although it is still lacks effectiveness, de novo design is expected to be developedinto a routine tool for drug discovery and attract wider acceptance by medicinal chemists. Amajor weakness of de novo design is its poor ability in scoring and the prediction of bindingenergies. New scoring functions should help to achieve a balance of accuracy and speed. Takingthe advantages of both experimental and computational FBDD and combining them into anintegrated drug design process will maximize the efficiency of drug discovery. With the develop-ment of both experimental and computational FBDD technologies to address these challengingproblems, FBDD should become accessible to both academia and the pharmaceutical industry,and promote drug discovery at a faster pace.

ACKNOWLEDGMENTS

We gratefully acknowledge financial support from the National Natural Science Foun-dation of China (Grants 30930107 and 30973640), the 863 Hi-Tech Program of China(Grant 2012AA020302). Shanghai Municipal Health Bureau (Grant XYQ2011038), and KeyLaboratory of Drug Research for Special Environments, PLA.

REFERENCES

1. Mayr LM, Bojanic D. Novel trends in high-throughput screening. Curr Opin Pharmacol 2009;9:580–588.

2. Bohacek RS, McMartin C, Guida WC. The art and practice of structure-based drug design: Amolecular modeling perspective. Med Res Rev 1996;16:3–50.

3. Gribbon P, Sewing A. High-throughput drug discovery: What can we expect from HTS? DrugDiscov Today 2005;10:17–22.



4. Hajduk PJ, Greer J. A decade of fragment-based drug design: Strategic advances and lessons learned.Nat Rev Drug Discov 2007;6:211–219.

5. Shuker SB, Hajduk PJ, Meadows RP, Fesik SW. Discovering high-affinity ligands for proteins:SAR by NMR. Science 1996;274:1531–1534.

6. Chessari G, Woodhead AJ. From fragment to clinical candidate—A historical perspective. DrugDiscov Today 2009;14:668–675.

7. Schuffenhauer A, Ruedisser S, Marzinzik AL, Jahnke W, Blommers M, Selzer P, Jacoby E. Librarydesign for fragment based screening. Curr Top Med Chem 2005;5:751–762.

8. Siegal G, Ab E, Schultz J. Integration of fragment screening and library design. Drug Discov Today2007;12:1032–1039.

9. Lepre CA, Moore JM, Peng JW. Theory and applications of NMR-based screening in pharmaceu-tical research. Chem Rev 2004;104:3641–3676.

10. Swayze EE, Jefferson EA, Sannes-Lowery KA, Blyn LB, Risen LM, Arakawa S, Osgood SA,Hofstadler SA, Griffey RH. SAR by MS: A ligand based technique for drug lead discovery againststructured RNA targets. J Med Chem 2002;45:3816–3819.

11. Erlanson DA, Wells JA, Braisted AC. Tethering: Fragment-based drug discovery. Annu Rev BiophysBiomol Struct 2004;33:199–223.

12. Hartshorn MJ, Murray CW, Cleasby A, Frederickson M, Tickle IJ, Jhoti H. Fragment-based leaddiscovery using X-ray crystallography. J Med Chem 2005;48:403–413.

13. Danielson UH. Fragment library screening and lead characterization using SPR biosensors. CurrTop Med Chem 2009;9:1725–1735.

14. Neumann T, Junker HD, Schmidt K, Sekul R. SPR-based fragment screening: Advantages andapplications. Curr Top Med Chem 2007;7:1630–1642.

15. Rees DC, Congreve M, Murray CW, Carr R. Fragment-based lead discovery. Nat Rev Drug Discov2004;3:660–672.

16. Flaherty KT, Yasothan U, Kirkpatrick P. Vemurafenib. Nat Rev Drug Discov 2011;10:811–812.

17. Murray CW, Rees DC. The rise of fragment-based drug discovery. Nat Chem 2009;1:187–192.

18. Congreve M, Chessari G, Tisi D, Woodhead AJ. Recent developments in fragment-based drugdiscovery. J Med Chem 2008;51:3661–3680.

19. Zartler ER, Shapiro MJ. Fragonomics: Fragment-based drug discovery. Curr Opin Chem Biol2005;9:366–370.

20. Bembenek SD, Tounge BA, Reynolds CH. Ligand efficiency and fragment-based drug discovery.Drug Discov Today 2009;14:278–283.

21. Fruh V, Zhou Y, Chen D, Loch C, Ab E, Grinkova YN, Verheij H, Sligar SG, Bushweller JH, SiegalG. Application of fragment-based drug discovery to membrane proteins: Identification of ligandsof the integral membrane enzyme DsbB. Chem Biol 2010;17:881–891.

22. Bamborough P, Brown MJ, Christopher JA, Chung CW, Mellor GW. Selectivity of kinase inhibitorfragments. J Med Chem 2011;54:5131–5143.

23. Babaoglu K, Shoichet BK. Deconstructing fragment-based inhibitor discovery. Nat Chem Biol2006;2:720–723.

24. Warr WA. Fragment-based drug discovery: What really works. An interview with Sandy Farmer ofBoehringer Ingelheim. J Comput Aided Mol Des 2011;25:599–605.

25. Desjarlais RL. Using computational techniques in fragment-based drug discovery. Methods Enzy-mol 2011;493:137–155.

26. Gozalbes R, Carbajo RJ, Pineda-Lucena A. Contributions of computational chemistry and bio-physical techniques to fragment-based drug discovery. Curr Med Chem 2010;17:1769–1794.

27. Hoffer L, Renaud JP, Horvath D. Fragment-based drug design: Computational & experimentalstate of the art. Comb Chem High Throughput Screen 2011;14:500–520.

28. Hubbard RE, Chen I, Davis B. Informatics and modeling challenges in fragment-based drug dis-covery. Curr Opin Drug Discov Devel 2007;10:289–297.



29. Law R, Barker O, Barker JJ, Hesterkamp T, Godemann R, Andersen O, Fryatt T, Courtney S,Hallett D, Whittaker M. The multiple roles of computational chemistry in fragment-based drugdesign. J Comput Aided Mol Des 2009;23:459–473.

30. Vangrevelinghe E, Rudisser S. Computational approaches for fragment optimization. Curr Comput-Aided Drug Design 2007;3:69–83.

31. Villar HO, Hansen MR. Computational techniques in fragment based drug discovery. Curr TopMed Chem 2007;7:1509–1513.

32. Zoete V, Grosdidier A, Michielin O. Docking, virtual high throughput screening and in silicofragment-based drug design. J Cell Mol Med 2009;13:238–248.

33. Makara GM. On sampling of fragment space. J Med Chem 2007;50:3214–3221.

34. Delaney JS. Predicting aqueous solubility from structure. Drug Discov Today 2005;10:289–295.

35. Faller B, Ertl P. Computational approaches to determine drug solubility. Adv Drug Deliv Rev2007;59(7):533–545.

36. Fejzo J, Lepre CA, Peng JW, Bemis GW, Ajay, Murcko MA, Moore JM. The SHAPES strat-egy: An NMR-based approach for lead generation in drug discovery. Chem Biol 1999;6:755–769.

37. Lepre C. Fragment-based drug discovery using the SHAPES method. Expert Opin Drug Discov2007;2:1555–1566.

38. Chung S, Parker JB, Bianchet M, Amzel LM, Stivers JT. Impact of linker strain and flexibility inthe design of a fragment-based inhibitor. Nat Chem Biol 2009;5:407–413.

39. Schneider G, Fechner U. Computer-based de novo design of drug-like molecules. Nat Rev DrugDiscov 2005;4:649–663.

40. Zhu Z, Sun ZY, Ye Y, Voigt J, Strickland C, Smith EM, Cumming J, Wang L, Wong J, Wang YS,Wyss DF, Chen X, Kuvelkar R, Kennedy ME, Favreau L, Parker E, McKittrick BA, StamfordA, Czarniecki M, Greenlee W, Hunter JC. Discovery of cyclic acylguanidines as highly potentand selective beta-site amyloid cleaving enzyme (BACE) inhibitors: Part I—Inhibitor design andvalidation. J Med Chem 2010;53:951–965.

41. Johnson MC, Hu Q, Lingardo L, Ferre RA, Greasley S, Yan J, Kath J, Chen P, Ermolieff J, Alton G.Novel isoquinolone PDK1 inhibitors discovered through fragment-based lead discovery. J ComputAided Mol Des 2011;25:689–698.

42. Congreve M, Carr R, Murray C, Jhoti H. A ‘rule of three’ for fragment-based lead discovery? DrugDiscov Today 2003;8:876–877.

43. Card GL, Blasdel L, England BP, Zhang C, Suzuki Y, Gillette S, Fong D, Ibrahim PN, Ar-tis DR, Bollag G, Milburn MV, Kim SH, Schlessinger J, Zhang KY. A family of phosphodi-esterase inhibitors discovered by cocrystallography and scaffold-based drug design. Nat Biotechnol2005;23:201–207.

44. Davies DR, Mamat B, Magnusson OT, Christensen J, Haraldsson MH, Mishra R, Pease B, HansenE, Singh J, Zembower D, Kim H, Kiselyov AS, Burgin AB, Gurney ME, Stewart LJ. Discoveryof leukotriene A4 hydrolase inhibitors using metabolomics biased fragment crystallography. J MedChem 2009;52:4694–4715.

45. Law RJ. Tetrabromobisphenol A: Investigating the worst-case scenario. Mar Pollut Bull2009;58(4):459–460.

46. Bondensgaard K, Ankersen M, Thogersen H, Hansen BS, Wulff BS, Bywater RP. Recognition ofprivileged structures by G-protein coupled receptors. J Med Chem 2004;47:888–899.

47. Schnur DM, Hermsmeier MA, Tebben AJ. Are target-family-privileged substructures truly privi-leged? J Med Chem 2006;49:2000–2009.

48. Clark M, Wiseman JS. Fragment-based prediction of the clinical occurrence of long QT syndromeand torsade de pointes. J Chem Inf Model 2009;49:2617–2626.

49. PubChem database. http://www.pubchem.ncbi.nlm.nih.gov.

50. eMolecules database. http://www.emolecules.com.



51. Oprea TI, Blaney JM. Cheminformatics approaches to fragment-based lead discovery. In: JahnkeW, Erlanson DA, Eds. Fragment-Based Approaches in Drug Discovery. Methods and Principles inMedicinal Chemistry, Vol. 34. Weinheim: Wiley-VCH Verlag GmbH; 2006. pp 91–111.

52. Tanaka N, Ohno K, Niimi T, Moritomo A, Mori K, Orita M. Small-world phenomena in chemicallibrary networks: Application to fragment-based drug discovery. J Chem Inf Model 2009;49:2677–2686.

53. World Drug Index; Daylight Chemical Information Systems, Inc. PO Box 7737, Laguna Niguel, CA92677.

54. MedChem03 database. BioByte: Claremont CA, and Daylight Chemical Information Systems, Inc.:Aliso Viejo, CA.

55. MDL Drug Data Report. Symyx Technologies, Inc.: Sunnyvale, CA.

56. Chen H, Gao J, Lu Y, Kou G, Zhang H, Fan L, Sun Z, Guo Y, Zhong Y. Preparation andcharacterization of PE38KDEL-loaded anti-HER2 nanoparticles for targeted cancer therapy.J Control Release 2008;128:209–216.

57. Horst EVD, IJzerman AP. Computational approaches to fragment and substructure discoveryand evaluation. In: Zartler ER, Shapiro MJ, Eds. Fragment-Based Drug Discovery: A PracticalApproach. John Wiley & Sons, Ltd. The Atrium, Southern Gate, Chichester, West Sussex, PO198SQ, United Kingdom; 2008. pp 199–222.

58. Zhang M, Sheng C, Xu H, Song Y, Zhang W. Constructing virtual combinatorial fragment librariesbased upon MDL Drug Data Report database. Sci China Ser B 2007;50:364–371.

59. Lewell XQ, Judd DB, Watson SP, Hann MM. RECAP—Retrosynthetic combinatorial analysisprocedure: A powerful new technique for identifying privileged molecular fragments with usefulapplications in combinatorial chemistry. J Chem Inf Comput Sci 1998;38:511–522.

60. Bemis GW, Murcko MA. Properties of known drugs. 2. Side chains. J Med Chem 1999;42:5095–5099.

61. Bemis GW, Murcko MA. The properties of known drugs. 1. Molecular frameworks. J Med Chem1996;39:2887–2893.

62. Hartenfeller M, Schneider G. De novo drug design. Methods Mol Biol 2011;672:299–323.

63. Schulz MN, Landstrom J, Bright K, Hubbard RE. Design of a fragment library that maximallyrepresents available chemical space. J Comput Aided Mol Des 2011;25:611–620.

64. Rarey M, Stahl M. Similarity searching in large combinatorial chemistry spaces. J Comput AidedMol Des 2001;15:497–520.

65. Hann MM, Oprea TI. Pursuing the leadlikeness concept in pharmaceutical research. Curr OpinChem Biol 2004;8:255–263.

66. Carr RA, Congreve M, Murray CW, Rees DC. Fragment-based lead discovery: Leads by design.Drug Discov Today 2005;10:987–992.

67. Mauser H, Stahl M. Chemical fragment spaces for de novo design. J Chem Inf Model 2007;47:318–324.

68. van Deursen R, Blum LC, Reymond JL. Visualisation of the chemical space of fragments, lead-likeand drug-like molecules in PubChem. J Comput Aided Mol Des 2011;25:649–662.

69. Andrews KM, Cramer RD. Toward general methods of targeted library design: Topomer shapesimilarity searching with diverse structures as queries. J Med Chem 2000;43:1723–1740.

70. Cramer RD, Soltanshahi F, Jilek R, Campbell B. AllChem: Generating and searching 10(20)synthetically accessible structures. J Comput Aided Mol Des 2007;21:341–350.

71. Jilik RJ, Cramer RD. Topomers: A validated protocol for their self-consistent generation. J ChemInf Comput Sci 2004;44:1121–1127.

72. Rarey M, Dixon JS. Feature trees: A new molecular similarity measure based on tree matching. JComput Aided Mol Des 1998;12:471–490.

73. Lessel U, Wellenzohn B, Lilienthal M, Claussen H. Searching fragment spaces with feature trees. JChem Inf Model 2009;49:270–279.



74. Lee ML, Schneider G. Scaffold architecture and pharmacophoric properties of natural productsand trade drugs: Application in the design of natural product-based combinatorial libraries. J CombChem 2001;3:284–289.

75. Siegel MG, Vieth M. Drugs in other drugs: A new look at drugs as fragments. Drug Discov Today2007;12:71–79.

76. Sutherland JJ, Higgs RE, Watson I, Vieth M. Chemical fragments as foundations for understandingtarget space and activity prediction. J Med Chem 2008;51:2689–2700.

77. Grabowski K, Schneider G. Properties and architecture of drugs and natural products revisited.Curr Chem Biol 2007;1:115–127.

78. Wang J, Hou T. Drug and drug candidate building block analysis. J Chem Inf Model 2010;50:55–67.

79. Vieth M, Siegel M. Structural fragments in marketed oral drugs. In: Jahnke W, Erlanson DA,Eds. Fragment-Based Approaches in Drug Discovery. Weinheim: Wiley-VCH Verlag GmbH & Co.KGaA; 2006. pp 113–124.

80. Morphy R, Rankovic Z. Designed multiple ligands. An emerging drug discovery paradigm. J MedChem 2005;48:6523–6543.

81. Sheridan RP. Finding multiactivity substructures by mining databases of drug-like compounds. JChem Inf Comput Sci 2003;43:1037–1050.

82. Sheridan RP. The most common chemical replacements in drug-like compounds. J Chem InfComput Sci 2002;42:103–108.

83. Ertl P. Cheminformatics analysis of organic substituents: Identification of the most common sub-stituents, calculation of substituent properties, and automatic identification of drug-like bioisostericgroups. J Chem Inf Comput Sci 2003;43:374–380.

84. Haubertin DY, Bruneau P. A database of historically-observed chemical replacements. J Chem InfModel 2007;47:1294–1302.

85. Kho R, Hodges JA, Hansen MR, Villar HO. Ring systems in mutagenicity databases. J Med Chem2005;48:6671–6678.

86. Kazius J, Nijssen S, Kok J, Back T, Ijzerman AP. Substructure mining using elaborate chemicalrepresentation. J Chem Inf Model 2006;46:597–605.

87. Batista J, Godden JW, Bajorath J. Assessment of molecular similarity from the analysis of randomlygenerated structural fragment populations. J Chem Inf Model 2006;46:1937–1944.

88. Batista J, Bajorath J. Chemical database mining through entropy-based molecular similarity assess-ment of randomly generated structural fragment populations. J Chem Inf Model 2007;47:59–68.

89. Batista J, Bajorath J. Mining of randomly generated molecular fragment populations uncoversactivity-specific fragment hierarchies. J Chem Inf Model 2007;47:1405–1413.

90. Lounkine E, Auer J, Bajorath J. Formal concept analysis for the identification of molecular fragmentcombinations specific for active and highly potent compounds. J Med Chem 2008;51:5342–5348.

91. Lameijer EW, Kok JN, Back T, Ijzerman AP. Mining a chemical database for fragment co-occurrence: Discovery of “chemical cliches”. J Chem Inf Model 2006;46:553–562.

92. Wilkens SJ, Janes J, Su AI. HierS: Hierarchical scaffold clustering using topological chemical graphs.J Med Chem 2005;48:3182–3193.

93. Schuffenhauer A, Ertl P, Roggo S, Wetzel S, Koch MA, Waldmann H. The scaffold tree—Visualization of the scaffold universe by hierarchical scaffold classification. J Chem Inf Model2007;47:47–58.

94. Shelat AA, Guy RK. Scaffold composition and biological relevance of screening libraries. Nat ChemBiol 2007;3:442–446.

95. Clark AM, Labute P. Detection and assignment of common scaffolds in project databases of leadmolecules. J Med Chem 2009;52:469–483.

96. Clark AM. 2D depiction of fragment hierarchies. J Chem Inf Model 2010;50:37–46.

97. Agrafiotis DK, Wiener JJ. Scaffold explorer: An interactive tool for organizing and mining structure-activity data spanning multiple chemotypes. J Med Chem 2010;53:5002–5011.



98. Renner S, van Otterlo WA, Dominguez Seoane M, Mocklinghoff S, Hofmann B, Wetzel S, Schuffen-hauer A, Ertl P, Oprea TI, Steinhilber D, Brunsveld L, Rauh D, Waldmann H. Bioactivity-guidedmapping and navigation of chemical space. Nat Chem Biol 2009;5:585–592.

99. Wetzel S, Klein K, Renner S, Rauh D, Oprea TI, Mutzel P, Waldmann H. Interactive explorationof chemical space with Scaffold Hunter. Nat Chem Biol 2009;5:581–583.

100. Hopkins AL, Groom CR, Alex A. Ligand efficiency: A useful metric for lead selection. Drug DiscovToday 2004;9:430–431.

101. Mocklinghoff S, van Otterlo WA, Rose R, Fuchs S, Zimmermann TJ, Dominguez Seoane M,Waldmann H, Ottmann C, Brunsveld L. Design and evaluation of fragment-like estrogen receptortetrahydroisoquinoline ligands from a scaffold-detection approach. J Med Chem 2011;54:2005–2011.

102. Mattos C, Bellamacina CR, Peisach E, Pereira A, Vitkup D, Petsko GA, Ringe D. Multiple sol-vent crystal structures: Probing binding sites, plasticity and hydration. J Mol Biol 2006;357:1471–1482.

103. Mattos C, Ringe D. Locating and characterizing binding sites on proteins. Nat Biotechnol1996;14:595–599.

104. Leis S, Schneider S, Zacharias M. In silico prediction of binding sites on proteins. Curr Med Chem2010;17:1550–1562.

105. Laurie AT, Jackson RM. Methods for the prediction of protein-ligand binding sites for structure-based drug design and virtual ligand screening. Curr Protein Pept Sci 2006;7:395–406.

106. Goodford PJ. A computational procedure for determining energetically favorable binding sites onbiologically important macromolecules. J Med Chem 1985;28:849–857.

107. von Itzstein M, Wu WY, Kok GB, Pegg MS, Dyason JC, Jin B, Van Phan T, Smythe ML, White HF,Oliver SW, Colman PM, Varghese JN, Ryan DM, Woods JM, Bethell RC, Hotham VJ, CameronJM, Penn CR. Rational design of potent sialidase-based inhibitors of influenza virus replication.Nature 1993;363:418–423.

108. Miranker A, Karplus M. Functionality maps of binding sites: A multiple copy simultaneous searchmethod. Proteins 1991;11:29–34.

109. Schubert C, Stultz C. The multi-copy simultaneous search methodology: A fundamental tool forstructure-based drug design. J Comput Aided Mol Des 2009;23:475–489.

110. Campbell SJ, Gold ND, Jackson RM, Westhead DR. Ligand binding: Functional site location,similarity and docking. Curr Opin Struct Biol 2003;13:389–395.

111. Sotriffer C, Klebe G. Identification and mapping of small-molecule binding sites in proteins: Com-putational tools for structure-based drug design. Farmaco 2002;57:243–251.

112. Landon M, Lieberman R, Hoang Q, Ju S, Caaveiro J, Orwig S, Kozakov D, Brenke R, Chuang G,Beglov D, Vajda S, Petsko G, Ringe D. Detection of ligand binding hot spots on protein surfacesvia fragment-based methods: Application to DJ-1 and glucocerebrosidase. J Comput Aided MolDes 2009;23:491–500.

113. Clark M, Guarnieri F, Shkurko I, Wiseman J. Grand canonical Monte Carlo simulation of ligand-protein binding. J Chem Inf Model 2006;46:231–242.

114. Irwin JJ, Shoichet BK. ZINC—A free database of commercially available compounds for virtualscreening. J Chem Inf Model 2005;45:177–182.

115. Nayal M, Honig B. On the nature of cavities on protein surfaces: Application to the identificationof drug-binding sites. Proteins 2006;63:892–906.

116. Leach AR, Shoichet BK, Peishoff CE. Prediction of protein-ligand interactions. Docking andscoring: Successes and gaps. J Med Chem 2006;49:5851–5855.

117. Teotico DG, Babaoglu K, Rocklin GJ, Ferreira RS, Giannetti AM, Shoichet BK. Docking forfragment inhibitors of AmpC beta-lactamase. Proc Natl Acad Sci USA 2009;106:7455–7460.

118. Ferrara P, Gohlke H, Price DJ, Klebe G, Brooks CL, 3rd. Assessing scoring functions for protein-ligand interactions. J Med Chem 2004;47:3032–3047.



119. Majeux N, Scarsi M, Caflisch A. Efficient electrostatic solvation model for protein-fragment dock-ing. Proteins 2001;42:256–268.

120. Friesner RA, Banks JL, Murphy RB, Halgren TA, Klicic JJ, Mainz DT, Repasky MP, Knoll EH,Shelley M, Perry JK, Shaw DE, Francis P, Shenkin PS. Glide: A new approach for rapid, accuratedocking and scoring. 1. Method and assessment of docking accuracy. J Med Chem 2004;47:1739–1749.

121. Halgren TA, Murphy RB, Friesner RA, Beard HS, Frye LL, Pollard WT, Banks JL. Glide: A newapproach for rapid, accurate docking and scoring. 2. Enrichment factors in database screening. JMed Chem 2004;47:1750–1759.

122. Kawatkar S, Wang H, Czerminski R, Joseph-McCarthy D. Virtual fragment screening: An explo-ration of various docking and scoring protocols for fragments using Glide. J Comput Aided MolDes 2009 23:527–539.

123. Sandor M, Kiss R, Keseru GM. Virtual fragment docking by Glide: A validation study on 190protein-fragment complexes. J Chem Inf Model 2010;50:1165–1172.

124. Verdonk ML, Giangreco I, Hall RJ, Korb O, Mortenson PN, Murray CW. Docking performanceof fragments and druglike compounds. J Med Chem 2011;54:5422–5431.

125. Knehans T, Schuller A, Doan DN, Nacro K, Hill J, Guntert P, Madhusudhan MS, Weil T, Va-sudevan SG. Structure-guided fragment-based in silico drug design of dengue protease inhibitors. JComput Aided Mol Des 2011;25:263–274.

126. Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS, Olson AJ. AutoDock4and AutoDockTools4: Automated docking with selective receptor flexibility. J Comput Chem2009;30:2785–2791.

127. Al-qattan MN, Mordi MN. Site-directed fragment-based generation of virtual sialic acid databasesagainst influenza A hemagglutinin. J Mol Model 2010;16:975–991.

128. Vadivelan S, Sinha BN, Tajne S, Jagarlapudi SA. Fragment and knowledge-based design of selectiveGSK-3beta inhibitors using virtual screening models. Eur J Med Chem 2009;44:2361–2371.

129. Brough PA, Barril X, Borgognoni J, Chene P, Davies NG, Davis B, Drysdale MJ, Dymock B, EcclesSA, Garcia-Echeverria C, Fromont C, Hayes A, Hubbard RE, Jordan AM, Jensen MR, MasseyA, Merrett A, Padfield A, Parsons R, Radimerski T, Raynaud FI, Robertson A, Roughley SD,Schoepfer J, Simmonite H, Sharp SY, Surgenor A, Valenti M, Walls S, Webb P, Wood M, WorkmanP, Wright L. Combining hit identification strategies: Fragment-based and in silico approaches toorally active 2-aminothieno[2,3-d]pyrimidine inhibitors of the Hsp90 molecular chaperone. J MedChem 2009;52:4794–4809.

130. Gleeson MP, Gleeson D. QM/MM as a tool in fragment based drug discovery. A cross-docking,rescoring study of kinase inhibitors. J Chem Inf Model 2009;49:1437–1448.

131. Marcou G, Rognan D. Optimizing fragment and scaffold docking by use of molecular interactionfingerprints. J Chem Inf Model 2007;47:195–207.

132. Fukunishi Y, Mashimo T, Orita M, Ohno K, Nakamura H. In silico fragment screening by replicageneration (FSRG) method for fragment-based drug design. J Chem Inf Model 2009;49:925–933.

133. Li H, Li C. Multiple ligand simultaneous docking: Orchestrated dancing of ligands in binding sitesof protein. J Comput Chem 2010;31:2014–2022.

134. Li H, Liu A, Zhao Z, Xu Y, Lin J, Jou D, Li C. Fragment-based drug design and drug reposi-tioning using multiple ligand simultaneous docking (MLSD): Identifying celecoxib and templatecompounds as novel inhibitors of signal transducer and activator of transcription 3 (STAT3). J MedChem 2011;54:5592–5596.

135. Ekonomiuk D, Su XC, Ozawa K, Bodenreider C, Lim SP, Otting G, Huang D, Caflisch A. Flaviviralprotease inhibitors identified by fragment-based library docking into a structure generated bymolecular dynamics. J Med Chem 2009;52:4860–4868.

136. Danziger DJ, Dean PM. Automated site-directed drug design: A general algorithm for knowledgeacquisition about hydrogen-bonding regions at protein surfaces. Proc R Soc Lond B Biol Sci1989;236:101–113.



137. Clark DE, Westhead DR. Evolutionary algorithms in computer-aided molecular design. J ComputAided Mol Des 1996;10:337–358.

138. Dorigo M, Di Caro G, Gambardella LM. Ant algorithms for discrete optimization. Artif Life1999;5:137–172.

139. Hiss JA, Hartenfeller M, Schneider G. Concepts and applications of “natural computing” techniquesin de novo drug and peptide design. Curr Pharm Des 2010;16:1656–1665.

140. DeWitt R, Shaknovich E. SmoG: De novo design method based on simple, fast, and accurate freeenergy estimates. 1. Methodology and supporting evidence. J Am Chem Soc 1996;118:11733–11744.

141. DeWitt R, Shaknovich E. SmoG: De novo design method based on simple, fast, and accurate freeenergy estimates. 2. Case studies on molecular design. J Am Chem Soc 1997;119:4608–4617.

142. Bohacek RS, McMartin C. Multiple highly diverse structures complementary to enzyme bindingsites: Results of extensive application of de novo design method incorporating combinatorial growth.J Am Chem Soc 1994;116:5560–5571.

143. Rotstein SH, Murcko MA. GroupBuild: A fragment-based method for de novo drug design. J MedChem 1993;36:1700–1710.

144. Gillet VJ, Newell W, Mata P, Myatt G, Sike S, Zsoldos Z, Johnson AP. SPROUT: Recent develop-ments in the de novo design of molecules. J Chem Inf Comput Sci 1994;34:207–217.

145. Moon JB, Howe WJ. Computer design of bioactive molecules: A method for receptor-based denovo ligand design. Proteins 1991;11:314–328.

146. Rarey M, Kramer B, Lengauer T, Klebe G. A fast flexible docking method using an incrementalconstruction algorithm. J Mol Biol 1996;261:470–489.

147. Degen J, Rarey M. FlexNovo: Structure-based searching in large fragment spaces. ChemMedChem2006;1:854–868.

148. Zaliani A, Boda K, Seidel T, Herwig A, Schwab CH, Gasteiger J, Claußen H, Lemmen C, Degen J,Parn J, Rarey M. Second-generation de novo design: A view from a medicinal chemist perspective.J Comput Aided Mol Des 2009;23:593–602.

149. Boehm M, Wu TY, Claussen H, Lemmen C. Similarity searching and scaffold hopping in syntheti-cally accessible combinatorial chemistry spaces. J Med Chem 2008;51:2468–2480.

150. Parn J, Degen J, Rarey M. Exploring fragment spaces under multiple physicochemical constraints.J Comput Aided Mol Des 2007;21:327–340.

151. Durrant JD, Amaro RE, McCammon JA. AutoGrow: A novel algorithm for protein inhibitordesign. Chem Biol Drug Des 2009;73:168–178.

152. Morris GM, Goodsell DS, Halliday RS, Huey R, Hart WE, Belew RK, Olson AJ. Automateddocking using a Lamarckian genetic algorithm and an empirical binding free energy function. JComput Chem 1998;19:1639–1662.

153. Kutchukian PS, Lou D, Shakhnovich EI. FOG: Fragment Optimized Growth algorithm for the denovo generation of molecules occupying druglike chemical space. J Chem Inf Model 2009;49:1630–1642.

154. Pearlman DA, Murcko MA. CONCERTS: Dynamic connection of fragments as an approach to denovo ligand design. J Med Chem 1996;39:1651–1663.

155. Bohm HJ. The computer program LUDI: A new method for the de novo design of enzyme inhibitors.J Comput Aided Mol Des 1992;6:61–78.

156. Bohm HJ. On the use of LUDI to search the Fine Chemicals Directory for ligands of proteins ofknown three-dimensional structure. J Comput Aided Mol Des 1994;8:623–632.

157. Lauri G, Bartlett PA. CAVEAT: A program to facilitate the design of organic molecules. J ComputAided Mol Des 1994;8:51–66.

158. Tschinke V, Cohen NC. The NEWLEAD program: A new method for the design of candidatestructures from pharmacophoric hypotheses. J Med Chem 1993;36:3863–3870.

159. Miranker A, Karplus M. An automated method for dynamic ligand design. Proteins 1995;23:472–490.



160. Roe DC, Kuntz ID. BUILDER v.2: Improving the chemistry of a de novo design strategy. J ComputAided Mol Des 1995;9:269–282.

161. Stahl M, Todorov NP, James T, Mauser H, Boehm HJ, Dean PM. A validation study on thepractical use of automated de novo design. J Comput Aided Mol Des 2002;16:459–478.

162. Dey F, Caflisch A. Fragment-based de novo ligand design by multiobjective evolutionary optimiza-tion. J Chem Inf Model 2008;48:679–690.

163. Majeux N, Scarsi M, Apostolakis J, Ehrhardt C, Caflisch A. Exhaustive docking of molecularfragments with electrostatic solvation. Proteins 1999;37:88–105.

164. Ji H, Stanton BZ, Igarashi J, Li H, Martasek P, Roman LJ, Poulos TL, Silverman RB. Minimalpharmacophoric elements and fragment hopping, an approach directed at molecular diversity andisozyme selectivity. Design of selective neuronal nitric oxide synthase inhibitors. J Am Chem Soc2008;130:3900–3914.

165. Ji H, Li H, Martasek P, Roman LJ, Poulos TL, Silverman RB. Discovery of highly potent andselective inhibitors of neuronal nitric oxide synthase by fragment hopping. J Med Chem 2009;52:779–797.

166. Yuan Y, Pei J, Lai L. LigBuilder 2: A practical de novo drug design approach. J Chem Inf Model2011;51:1083–1091.

167. Wang R, Gao Y, Lai L. A multi-purpose program for structure-based drug design. J Mol Model2000;6:498–516.

168. Boda K, Seidel T, Gasteiger J. Structure and reaction based evaluation of synthetic accessibility.J Comput Aided Mol Des 2007;21:311–325.

169. Meunier B. Hybrid molecules with a dual mode of action: Dream or reality? Acc Chem Res2008;41:69–77.

170. Viegas-Junior C, Danuello A, da Silva Bolzani V, Barreiro EJ, Fraga CA. Molecular hybridization:A useful tool in the design of new drug prototypes. Curr Med Chem 2007;14:1829–1852.

171. Pierce AC, Rao G, Bemis GW. BREED: Generating novel inhibitors through hybridization ofknown ligands. Application to CDK2, p38, and HIV protease. J Med Chem 2004;47:2768–2775.

172. Li Y, Zhao Y, Liu Z, Wang R. Automatic tailoring and transplanting: A practical method thatmakes virtual screening more useful. J Chem Inf Model 2011;51:1474–1491.

173. Nisius B, Rester U. Fragment shuffling: An automated workflow for three-dimensional fragment-based ligand design. J Chem Inf Model 2009;49:1211–1222.

174. Moriaud F, Doppelt-Azeroual O, Martin L, Oguievetskaia K, Koch K, Vorotyntsev A, AdcockSA, Delfaud F. Computational fragment-based approach at PDB scale by protein local similarity.J Chem Inf Model 2009;49:280–294.

175. Doppelt O, Moriaud F, Bornot A, de Brevern AG. Functional annotation strategy for proteinstructures. Bioinformation 2007;1:357–359.

176. Jambon M, Andrieu O, Combet C, Deleage G, Delfaud F, Geourjon C. The SuMo server: 3Dsearch for protein functional sites. Bioinformatics 2005;21:3929–3930.

177. Schneider G, Neidhart W, Giller T, Schmid G. “Scaffold-Hopping” by topological pharmacophoresearch: A contribution to virtual screening. Angew Chem Int Ed Engl 1999;38:2894–2896.

178. Krueger BA, Dietrich A, Baringhaus KH, Schneider G. Scaffold-hopping potential of fragment-based de novo design: The chances and limits of variation. Comb Chem High Throughput Screen2009;12:383–396.

179. Maass P, Schulz-Gasch T, Stahl M, Rarey M. Recore: A fast and versatile method for scaffoldhopping based on small molecule crystal structure conformations. J Chem Inf Model 2007;47:390–399.

180. Schneider G, Hartenfeller M, Reutlinger M, Tanrikulu Y, Proschak E, Schneider P. Voyages to the(un)known: Adaptive design of bioactive compounds. Trends Biotechnol 2009;27:18–26.

181. Hartenfeller M, Proschak E, Schuller A, Schneider G. Concept of combinatorial de novo design ofdrug-like molecules by particle swarm optimization. Chem Biol Drug Des 2008;72:16–26.



182. Fechner U, Schneider G. Flux (1): A virtual synthesis scheme for fragment-based de novo design. JChem Inf Model 2006;46:699–707.

183. Fechner U, Schneider G. Flux (2): Comparison of molecular mutation and crossover operators forligand-based de novo design. J Chem Inf Model 2007;47:656–667.

184. Proschak E, Zettl H, Tanrikulu Y, Weisel M, Kriegl JM, Rau O, Schubert-Zsilavecz M, Schneider G.From molecular shape to potent bioactive agents I: Bioisosteric replacement of molecular fragments.ChemMedChem 2009;4:41–44.

185. Proschak E, Sander K, Zettl H, Tanrikulu Y, Rau O, Schneider P, Schubert-Zsilavecz M, Stark H,Schneider G. From molecular shape to potent bioactive agents II: Fragment-based de novo design.ChemMedChem 2009;4:45–48.

186. EA-InVentor; Tripos International: St. Louis, MO. http://www.tripos.com.

187. Damewood JR, Jr, Lerman CL, Masek BB. NovoFLAP: A ligand-based de novo design approachfor the generation of medicinally relevant ideas. J Chem Inf Model 2010;50:1296–1303.

188. Feher M, Gao Y, Baber JC, Shirley WA, Saunders J. The use of ligand-based de novo design forscaffold hopping and sidechain optimization: Two case studies. Bioorg Med Chem 2008;16:422–427.

189. Liu Q, Masek B, Smith K, Smith J. Tagged fragment method for evolutionary structure-based denovo lead generation and optimization. J Med Chem 2007;50:5392–5402.

190. Masek BB, Shen L, Smith KM, Pearlman RS. Sharing chemical information without sharingchemical structure. J Chem Inf Model 2008;48:256–261.

191. Pham TA, Jain AN. Parameter estimation for scoring protein-ligand interactions using negativetraining data. J Med Chem 2006;49:5856–5868.

192. Jain AN. Surflex: Fully automatic flexible molecular docking using a molecular similarity-basedsearch engine. J Med Chem 2003;46:499–511.

193. Huang Q, Li LL, Yang SY. PhDD: A new pharmacophore-based de novo design method of drug-likemolecules combined with assessment of synthetic accessibility. J Mol Graph Model 2010;28:775–787.

194. Vieth M, Erickson J, Wang J, Webster Y, Mader M, Higgs R, Watson I. Kinase inhibitor datamodeling and de novo inhibitor design with fragment approaches. J Med Chem 2009;52:6456–6466.

195. Liew CY, Ma XH, Liu X, Yap CW. SVM model for virtual screening of Lck inhibitors. J Chem InfModel 2009;49:877–885.

196. Loving K, Alberts I, Sherman W. Computational approaches for fragment-based and de novodesign. Curr Top Med Chem 2010;10:14–32.

197. Massova I, Kollman PA. Computational alanine scanning to probe protein-protein interactions: Anovel approach to evaluate binding free energies. J Am Chem Soc 1999;121:8133–8143.

198. Still WC, Tempczyk A, Hawley RC, Hendrickson T. Semianalytical treatment of solvation formolecular mechanics and dynamics. J Am Chem Soc 1990;112:6127–6129.

199. Kim JT, Hamilton AD, Bailey CM, Domaoal RA, Wang L, Anderson KS, Jorgensen WL. FEP-guided selection of bicyclic heterocycles in lead optimization for non-nucleoside inhibitors of HIV-1reverse transcriptase. J Am Chem Soc 2006;128:15372–15373.

200. Kollman PA. Free energy calculations: Applications to chemical and biochemical phenomena. ChemRev 1993;93:2395–2417.

201. Oostenbrink C, van Gunsteren WF. Free energies of binding of polychlorinated biphenyls to theestrogen receptor from a single simulation. Proteins 2004;54:237–246.

202. van Gunsteren WF, Berendsen HJ. Thermodynamic cycle integration by computer simulation asa tool for obtaining free energy differences in molecular chemistry. J Comput Aided Mol Des1987;1:171–176.

203. Fogel GB. Computational intelligence approaches for pattern discovery in biological systems. BriefBioinform 2008;9:307–316.

204. Hecht D, Fogel GB. A novel in silico approach to drug discovery via computational intelligence.J Chem Inf Model 2009;49:1105–1121.



205. Nicolaou CA, Apostolakis J, Pattichis CS. De novo drug design using multiobjective evolutionarygraphs. J Chem Inf Model 2009;49:295–307.

206. Schneider G, Lee ML, Stahl M, Schneider P. De novo design of molecular architectures byevolutionary assembly of drug-derived building blocks. J Comput Aided Mol Des 2000;14:487–494.

207. Vinkers HM, de Jonge MR, Daeyaert FF, Heeres J, Koymans LM, van Lenthe JH, Lewi PJ,Timmerman H, Van Aken K, Janssen PA. SYNOPSIS: SYNthesize and OPtimize System in Silico.J Med Chem 2003;46:2765–2773.

208. Symyx Technology Inc., 2440 Camino Ramon, Suite 300, San Ramon, CA 94583, USA.

209. Law J, Zsoldos Z, Simon A, Reid D, Liu Y, Khew SY, Johnson AP, Major S, Wade RA, Ando HY.Route Designer: A retrosynthetic analysis tool utilizing automated retrosynthetic rule generation.J Chem Inf Model 2009;49:593–602.

210. Patel H, Bodkin MJ, Chen B, Gillet VJ. Knowledge-based approach to de novo design using reactionvectors. J Chem Inf Model 2009;49:1163–1184.

211. Reisen FH, Schneider G, Proschak E. Reaction-MQL: Line notation for functional transformation.J Chem Inf Model 2009;49:6–12.

212. Hartenfeller M. Development of a Computational Method for Reaction-Driven De Novo Designof Druglike Compounds. Frankfurt am Main: Goethe University; 2010. p 120.

213. Kutchukian PS, Shakhnovich EI. De novo design: Balancing novelty and confined chemical space.Expert Opin Drug Discov 2010;5:789–812.

214. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experimental and computational approachesto estimate solubility and permeability in drug discovery and development settings. Adv Drug DelivRev 2001;46:3–26.

215. Hutter MC. In silico prediction of drug properties. Curr Med Chem 2009;16:189–202.

216. Agarwal AK, Johnson AP, Fishwick CWG. Synthesis of de novo designed small-molecule inhibitorsof bacterial RNA polymerase. Tetrahedron 2008;64:10049–10054.

217. Alig L, Alsenz J, Andjelkovic M, Bendels S, Benardeau A, Bleicher K, Bourson A, David-Pierson P,Guba W, Hildbrand S, Kube D, Lubbers T, Mayweg AV, Narquizian R, Neidhart W, NettekovenM, Plancher JM, Rocha C, Rogers-Evans M, Rover S, Schneider G, Taylor S, Waldmeier P. Benzo-dioxoles: Novel cannabinoid-1 receptor inverse agonists for the treatment of obesity. J Med Chem2008;51:2115–2127.

218. Bhurruth-Alcor Y, Rost T, Jorgensen MR, Kontogiorgis C, Skorve J, Cooper RG, Sheridan JM,Hamilton WD, Heal JR, Berge RK, Miller AD. Synthesis of novel PPARalpha/gamma dual agonistsas potential drugs for the treatment of the metabolic syndrome and diabetes type II designed usinga new de novo design program PROTOBUILD. Org Biomol Chem 2011;9:1169–1188.

219. Cogan DA, Aungst R, Breinlinger EC, Fadra T, Goldberg DR, Hao MH, Kroe R, Moss N,Pargellis C, Qian KC, Swinamer AD. Structure-based design and subsequent optimization of 2-tolyl-(1,2,3-triazol-1-yl-4-carboxamide) inhibitors of p38 MAP kinase. Bioorg Med Chem Lett2008;18:3251–3255.

220. Cumming J, Babu S, Huang Y, Carrol C, Chen X, Favreau L, Greenlee W, Guo T, Kennedy M,Kuvelkar R, Le T, Li G, McHugh N, Orth P, Ozgur L, Parker E, Saionz K, Stamford A, StricklandC, Tadesse D, Voigt J, Zhang L, Zhang Q. Piperazine sulfonamide BACE1 inhibitors: Design,synthesis, and in vivo characterization. Bioorg Med Chem Lett 2010;20:2837–2842.

221. Cumming JN, Le TX, Babu S, Carroll C, Chen X, Favreau L, Gaspari P, Guo T, Hobbs DW,Huang Y, Iserloh U, Kennedy ME, Kuvelkar R, Li G, Lowrie J, McHugh NA, Ozgur L, Pan J,Parker EM, Saionz K, Stamford AW, Strickland C, Tadesse D, Voigt J, Wang L, Wu Y, Zhang L,Zhang Q. Rational design of novel, potent piperazinone and imidazolidinone BACE1 inhibitors.Bioorg Med Chem Lett 2008;18:3236–3241.

222. Dong X, Zhang Z, Wen R, Shen J, Shen X, Jiang H. Structure-based de novo design, synthesis,and biological evaluation of the indole-based PPARgamma ligands (I). Bioorg Med Chem Lett2006;16:5913–5916.



223. Firth-Clark S, Willems HM, Williams A, Harris W. Generation and selection of novel estrogen recep-tor ligands using the de novo structure-based design tool, SkelGen. J Chem Inf Model 2006;46:642–647.

224. Hangeland JJ, Cheney DL, Friends TJ, Swartz S, Levesque PC, Rich AJ, Sun L, Bridal TR, AdamLP, Normandin DE, Murugesan N, Ewing WR. Design and SAR of selective T-type calciumchannel antagonists containing a biaryl sulfonamide core. Bioorg Med Chem Lett 2008;18:474–478.

225. Heikkila T, Thirumalairajan S, Davies M, Parsons MR, McConkey AG, Fishwick CW, JohnsonAP. The first de novo designed inhibitors of Plasmodium falciparum dihydroorotate dehydrogenase.Bioorg Med Chem Lett 2006;16:88–92.

226. Herschhorn A, Lerman L, Weitman M, Gleenberg IO, Nudelman A, Hizi A. De novo parallel design,synthesis and evaluation of inhibitors against the reverse transcriptase of human immunodeficiencyvirus type-1 and drug-resistant variants. J Med Chem 2007;50:2370–2384.

227. Kandil S, Biondaro S, Vlachakis D, Cummins AC, Coluccia A, Berry C, Leyssen P, Neyts J, BrancaleA. Discovery of a novel HCV helicase inhibitor by a de novo drug design approach. Bioorg MedChem Lett 2009;19:2935–2937.

228. Liu B, Joseph RW, Dorsey BD, Schiksnis RA, Northrop K, Bukhtiyarova M, Springman EB.Structure-based design of substituted biphenyl ethylene ethers as ligands binding in the hydrophobicpocket of gp41 and blocking the helical bundle formation. Bioorg Med Chem Lett 2009;19:5693–5697.

229. Moore WR, Jr. Maximizing discovery efficiency with a computationally driven fragment approach.Curr Opin Drug Discov Devel 2005;8:355–364.

230. Park H, Bahn YJ, Ryu SE. Structure-based de novo design and biochemical evaluation of novelCdc25 phosphatase inhibitors. Bioorg Med Chem Lett 2009;19:4330–4334.

231. Schuller A, Suhartono M, Fechner U, Tanrikulu Y, Breitung S, Scheffer U, Gobel MW, SchneiderG. The concept of template-based de novo design from drug-derived molecular fragments and itsapplication to TAR RNA. J Comput Aided Mol Des 2008;22:59–68.

232. Sova M, Cadez G, Turk S, Majce V, Polanc S, Batson S, Lloyd AJ, Roper DI, Fishwick CW, GobecS. Design and synthesis of new hydroxyethylamines as inhibitors of D-alanyl-D-lactate ligase (VanA)and D-alanyl-D-alanine ligase (DdlB). Bioorg Med Chem Lett 2009;19:1376–1379.

233. Zhao L, Huang W, Liu H, Wang L, Zhong W, Xiao J, Hu Y, Li S. FK506-binding protein ligands:Structure-based design, synthesis, and neurotrophic/neuroprotective properties of substituted 5,5-dimethyl-2-(4-thiazolidine)carboxylates. J Med Chem 2006;49:4059–4071.

234. Silverman RB. Design of selective neuronal nitric oxide synthase inhibitors for the prevention andtreatment of neurodegenerative diseases. Acc Chem Res 2009;42:439–451.

235. Crane BR, Arvai AS, Ghosh DK, Wu C, Getzoff ED, Stuehr DJ, Tainer JA. Structure of nitricoxide synthase oxygenase dimer with pterin and substrate. Science 1998;279:2121–2126.

236. Raman CS, Li H, Martasek P, Kral V, Masters BS, Poulos TL. Crystal structure of constitutiveendothelial nitric oxide synthase: A paradigm for pterin function involving a novel metal center.Cell 1998;95:939–950.

237. Fischmann TO, Hruza A, Niu XD, Fossetta JD, Lunn CA, Dolphin E, Prongay AJ, Reichert P,Lundell DJ, Narula SK, Weber PC. Structural characterization of nitric oxide synthase isoformsreveals striking active-site conservation. Nat Struct Biol 1999;6:233–242.

238. Li H, Shimizu H, Flinspach M, Jamal J, Yang W, Xian M, Cai T, Wen EZ, Jia Q, Wang PG, PoulosTL. The novel binding mode of N-alkyl-N’-hydroxyguanidine to neuronal nitric oxide synthaseprovides mechanistic insights into NO biosynthesis. Biochemistry 2002;41:13868–13875.

239. Ji H, Delker SL, Li H, Martasek P, Roman LJ, Poulos TL, Silverman RB. Exploration of theactive site of neuronal nitric oxide synthase by the design and synthesis of pyrrolidinomethyl2-aminopyridine derivatives. J Med Chem 2010;53:7804–7824.

240. Ji H, Tan S, Igarashi J, Li H, Derrick M, Martasek P, Roman LJ, Vasquez-Vivar J, Poulos TL,Silverman RB. Selective neuronal nitric oxide synthase inhibitors and the prevention of cerebralpalsy. Ann Neurol 2009;65:209–217.



241. Fischer G, Gallay P, Hopkins S. Cyclophilin inhibitors for the treatment of HCV infection. CurrOpin Investig Drugs 2010;11:911–918.

242. Obchoei S, Wongkhan S, Wongkham C, Li M, Yao Q, Chen C. Cyclophilin A: Potential functionsand therapeutic target for human cancer. Med Sci Monit 2009;15:221–232.

243. Satoh K, Shimokawa H, Berk BC. Cyclophilin A: Promising new target in cardiovascular therapy.Circ J 2010;74:2249–2256.

244. Guichou JF, Viaud J, Mettling C, Subra G, Lin YL, Chavanieu A. Structure-based design, synthesis,and biological evaluation of novel inhibitors of human cyclophilin A. J Med Chem 2006;49:900–910.

245. Ni S, Yuan Y, Huang J, Mao X, Lv M, Zhu J, Shen X, Pei J, Lai L, Jiang H, Li J. Discoveringpotent small molecule inhibitors of cyclophilin A using de novo drug design approach. J Med Chem2009;52:5295–5298.

246. Nugiel DA, Krumrine JR, Hill DC, Damewood JR, Jr, Bernstein PR, Sobotka-Briner CD, Liu J,Zacco A, Pierson ME. De novo design of a picomolar nonbasic 5-HT(1B) receptor antagonist. JMed Chem 2010;53:1876–1880.

247. Ji H, Zhang W, Zhou Y, Zhang M, Zhu J, Song Y, Lu J. A three-dimensional model of lanosterol14alpha-demethylase of Candida albicans and its interaction with azole antifungals. J Med Chem2000;43:2493–2505.

248. Sheng C, Miao Z, Ji H, Yao J, Wang W, Che X, Dong G, Lu J, Guo W, Zhang W. Three-dimensional model of lanosterol 14 alpha-demethylase from Cryptococcus neoformans: Active-sitecharacterization and insights into azole binding. Antimicrob Agents Chemother 2009;53:3487–3495.

249. Sheng C, Zhang W, Zhang M, Song Y, Ji H, Zhu J, Yao J, Yu J, Yang S, Zhou Y, Lu J. Homologymodeling of lanosterol 14alpha-demethylase of Candida albicans and Aspergillus fumigatus andinsights into the enzyme-substrate interactions. J Biomol Struct Dyn 2004;22:91–99.

250. Sheng C, Zhang W. New lead structures in antifungal drug discovery. Curr Med Chem 2011;18:733–766.

251. Sheng C, Zhang W, Ji H, Zhang M, Song Y, Xu H, Zhu J, Miao Z, Jiang Q, Yao J, Zhou Y, LuJ. Structure-based optimization of azole antifungal agents by CoMFA, CoMSIA, and moleculardocking. J Med Chem 2006;49:2512–2525.

252. Che X, Sheng C, Wang W, Cao Y, Xu Y, Ji H, Dong G, Miao Z, Yao J, Zhang W. Newazoles with potent antifungal activity: Design, synthesis and molecular docking. Eur J Med Chem2009;44(10):4218–4226.

253. Sheng C, Che X, Wang W, Wang S, Cao Y, Yao J, Miao Z, Zhang W. Structure-based design,synthesis, and antifungal activity of new triazole derivatives. Chem Biol Drug Des 2011;78:309–313.

254. Sheng C, Wang W, Che X, Dong G, Wang S, Ji H, Miao Z, Yao J, Zhang W. Improved modelof lanosterol 14alpha-demethylase by ligand-supported homology modeling: Validation by virtualscreening and azole optimization. ChemMedChem 2010;5:390–397.

255. Wang W, Sheng C, Che X, Ji H, Cao Y, Miao Z, Yao J, Zhang W. Discovery of highly potent novelantifungal azoles by structure-based rational design. Bioorg Med Chem Lett 2009;19:5965–5969.

256. Wang W, Sheng C, Che X, Ji H, Miao Z, Yao J, Zhang WN. Design, synthesis, and antifungal activityof novel conformationally restricted triazole derivatives. Arch Pharm (Weinheim) 2009;342:732–739.

257. Wang W, Wang S, Liu Y, Dong G, Cao Y, Miao Z, Yao J, Zhang W, Sheng C. Novel conformation-ally restricted triazole derivatives with potent antifungal activity. Eur J Med Chem 2010;45:6020–6026.

258. Xu Y, Sheng C, Wang W, Che X, Cao Y, Dong G, Wang S, Ji H, Miao Z, Yao J, Zhang W. Structure-based rational design, synthesis and antifungal activity of oxime-containing azole derivatives. BioorgMed Chem Lett 2010;20:2942–2945.

259. Gao S, Tao X, Sun L, Sheng C, Zhang W, Yun Y, Li J, Miao H, Chen W. An liquid chromatography-tandem mass spectrometry assay for determination of trace amount of new antifungal drug iod-iconazole in human plasma. J Chromatogr B Analyt Technol Biomed Life Sci 2009;877:382–386.



260. Sun N, Wen J, Lu G, Hong Z, Fan G, Wu Y, Sheng C, Zhang W. An ultra-fast LC method forthe determination of iodiconazole in microdialysis samples and its application in the calibration oflaboratory-made linear probes. J Pharm Biomed Anal 2010;51:248–251.

261. Wen J, Fan GR, Hong ZY, Chai YF, Yin XP, Wu YT, Sheng CQ, Zhang WN. High perfor-mance liquid chromatographic determination of a new antifungal compound, ADKZ in rat plasma.J Pharm Biomed Anal 2007;43:655–658.

262. Ji H, Zhang W, Zhang M, Kudo M, Aoyama Y, Yoshida Y, Sheng C, Song Y, Yang S, ZhouY, Lu J, Zhu J. Structure-based de novo design, synthesis, and biological evaluation of non-azoleinhibitors specific for lanosterol 14alpha-demethylase of fungi. J Med Chem 2003;46:474–485.

263. Sheng C, Che X, Wang W, Wang S, Cao Y, Yao J, Miao Z, Zhang W. Design and synthesis ofantifungal benzoheterocyclic derivatives by scaffold hopping. Eur J Med Chem 2011;46:1706–1712.

264. Tang H, Zheng C, Lv J, Wu J, Li Y, Yang H, Fu B, Li C, Zhou Y, Zhu J. Synthesis and anti-fungal activities in vitro of novel pyrazino [2,1-a] isoquinolin derivatives. Bioorg Med Chem Lett2010;20:979–982.

265. Tang H, Zheng CH, Zhu J, Fu BY, Zhou YJ, Lv JG. Design and synthesis of novel pyrazino[2,1-a]isoquinolin derivatives with potent antifungal activity. Arch Pharm (Weinheim) 2010;343:360–366.

266. Yao B, Ji H, Cao Y, Zhou Y, Zhu J, Lu J, Li Y, Chen J, Zheng C, Jiang Y, Liang R, Tang H. Synthesisand antifungal activities of novel 2-aminotetralin derivatives. J Med Chem 2007;50:5293–5300.

267. Roche O, Rodriguez Sarmiento RM. A new class of histamine H3 receptor antagonists derived fromligand based design. Bioorg Med Chem Lett 2007;17:3670–3675.

Chunquan Sheng received his bachelor degree in pharmacy (2000) and PhD degree in medicinalchemistry (2005) from Second Military Medical University. Dr. Sheng was appointed to thefaculty at the Department of Medicinal Chemistry, School of Pharmacy in 2005, where he ispresently an Associate Professor of Medicinal Chemistry and the Deputy Director of Departmentof Medicinal Chemistry. His research interests are mainly in computer-aided drug design andsynthetic medicinal chemistry.

Wannian Zhang received his bachelor degree in pharmacy (1968) and MS degree in medicinalchemistry (1981) from Second Military Medical University. He worked as a Professor of Medic-inal Chemistry and the Director of Department of Medicinal Chemistry, School of PharmacySecond Military Medical University in 1992. From 1994 to 2001, Prof. Zhang was appointed asthe Dean of School of Pharmacy. Now, Prof. Zhang is the Chief of The State’s Key Discipline ofMedicinal Chemistry, Second Military Medical University. His research interests are mainly inantifungal and antitumor drug design and development.


Med 21255

Documents

computational fbdd

highthroughput fragment

eachstep of experimental

computational tools

lead compounds

improved lead

highthroughput screening

successrate of lead